WO2022241246A1

WO2022241246A1 - Techniques for optimizing detection of particles captured by a microfluidic system

Info

Publication number: WO2022241246A1
Application number: PCT/US2022/029240
Authority: WO
Inventors: Monika WEBER; David FRAEBEL; Robert Weber; Slawomir ANTOSZCZYK; Fred FARBER; Elliot Swart
Original assignee: Fluid-Screen, Inc.
Priority date: 2021-05-13
Filing date: 2022-05-13
Publication date: 2022-11-17

Abstract

Methods and apparatus for detecting, classifying, and/or quantifying a particle in an image are provided. A method comprises determining whether an intensity value for the particle in the image is greater than a first threshold value, determining whether at least one morphological characteristic of the particle in the image satisfies a first set of criteria associated with particles in a first group of particles, classifying the particle as belonging to the first group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image satisfies the first set of criteria, and outputting a result of the classifying of the particle as belonging to the first group of particles.

Description

TECHNIQUES FOR OPTIMIZING DETECTION OF PARTICLES CAPTURED BY A

MICROFLUIDIC SYSTEM

RELATED APPLICATIONS

[0001] This application claims priority under 35 U.S.C. § 119(e) to U.S. Provisional Application No.: 63/188,347 filed May 13, 2021 and entitled TECHNIQUES FOR OPTIMIZING DETECTION OF PARTICLES CAPTURED BY A MICROFLUIDIC SYSTEM,” the entire contents of which is incorporated by reference herein.

BACKGROUND

[0002] Detection and identification of particles (e.g., bacterial and viral pathogens) present in cell containing solutions (e.g., blood, urine, CSF, mammalian cell culture, CHO cell matrix, CAR-T drug product, CAR-T specimen, CAR-NK drug product, body fluids, apheresis samples or samples related to immunotherapy), protein containing solutions (e.g., for pharmaceuticals during manufacturing, drug product, drug substance), analyte extraction from microbiome samples, water, sterile fluids and other fluids is possible by employing isolation on cultural media and metabolic fingerprinting methods. Isozyme analysis, direct colony thin layer chromatography and gel electrophoresis techniques have been successfully applied for the detection of some bacterial pathogens. Immunoassay and nucleic acid-based assays are now widely accepted techniques, providing more sensitive and specific detection and quantification of bacteria.

[0003] Dielectrophoresis (DEP) relates to a force in an electric field gradient on objects having dielectric moments. DEP has shown promise for particle separation, but has not yet been applied in clinical settings, pharmaceutical quality assurance settings, or immunotherapy. DEP uses a natural or induced dipole to cause a net force on a particle in a region having an electric field gradient. The force depends on the Clausius-Mossotti factor associated with the particle.

SUMMARY

[0004] In some embodiments, there is provided data analysis techniques, which differentiate between different particles (e.g., bacterial cells and debris) in an image. For example, the techniques may be performed by identifying one or more characteristics (e.g., intensity, size, shape) of the particles which may be used to classify the particle (e.g., using thresholds, ranges, to determine how to classify a particle). In some embodiments, the techniques may provide for removing particles classified as debris from a data set. In some embodiments, the techniques may provide for separating particles classified as debris into a separate data set.

[0005] In some embodiments, there is provided techniques for using computer vision to automatically detect and/or quantify microbes. In some embodiments, computer vision may be implemented to detect objects in an image. The objects may be classified as a microbe of interest or any other object. Such techniques may be compared with manual inspection to evaluate the accuracy of the computer vision techniques.

[0006] In some embodiments, a method may be provided for combining the results of multiple quantification techniques. For example, in some embodiments, two or more techniques (e.g., the same algorithm with two sets of parameters or two different algorithms) may be used to obtain two values for a quantification of microbes in a sample. The two values may be averaged to obtain an improved value for a quantification of microbes in a sample.

[0007] In some embodiments, a machine learning algorithm may be provided for quantification of microbes, for example, according to the techniques described herein.

[0008] In some embodiments, there is provided techniques for performing negative control scans. A negative control scan may indicate the presence of particles on an electrode surface in the absence of any sample. The negative control scan may be subtracted from data obtained during sample processing.

[0009] In some embodiments, there is provided techniques for removing nucleic acids from a sample prior to processing the sample with a microfluidic system. For example, in some embodiments there is provided techniques for enzymatically degrading nucleic acids (e.g., from burst cells) in a sample to prevent the nucleic acids from interfering with sample processing. In some embodiments, there is provided techniques for removing surfactants from a sample prior to processing the sample with a microfluidic system. For example, in some embodiments there is provided techniques for adsorbing surfactants in a sample to prevent the surfactants from interfering with sample processing.

[0010] In some embodiments, there is provided techniques for reducing the time and cost of sterility testing on a sample. For example, a sample may be first processed by a microfluidic system. The microfluidic system and/or techniques used to analyze the data obtained from the microfluidic system may be optimized to eliminate false negative results. Based on the data obtained from the microfluidic system, the sample may be determined to be ready for real time release, or further testing may be performed (e.g., bacterial cultures).

[0011] In some embodiments, a method for classifying a particle in an image as belonging to a first group of particles, the particle being represented in the image as a plurality of contiguous pixels of the image is provided. The method comprises determining whether an intensity value for the particle in the image is greater than a first threshold value, determining whether at least one morphological characteristic of the particle in the image satisfies a first set of criteria associated with particles in the first group of particles, classifying the particle as belonging to the first group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image satisfies the first set of criteria, and outputting a result of the classifying of the particle as belonging to the first group of particles.

[0012] In one aspect, the morphological characteristic comprises a size and/or shape of the particle. In one aspect, the size of the particle comprises a number of contiguous pixels representing the particle in the image. In one aspect, the size of the particle is determined based on at least one measurement selected from the group consisting of an area, a circumference, a diameter, an extent, and a bisector. In one aspect, the shape of the particle is determined based on at least one measurement selected from the group consisting of an aspect ratio, an elongation value, a convexity value, a shape factor, and a sphericity value. In one aspect, the particle comprises one of a microorganism or debris. In one aspect, the microorganism comprises one of bacteria, yeast, or mold. In one aspect, the first group of particles comprise microorganisms. [0013] In one aspect, the image includes particles belonging to the first group of particles and particles belonging to a second group of particles, and the method further comprises classifying the particle as belonging to the second group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image does not satisfy the first set of criteria. In one aspect, the second group of particles comprises debris. In one aspect, the second group of particles comprises components other than microorganisms.

[0014] In one aspect, the method further comprises when the particle is classified as belonging to the first group of particles, further classifying the particle as belonging to a first subgroup or a second subgroup of the first group of particles at least in part by determining whether the value for the at least one morphological characteristic satisfies a second set of criteria, and classifying the particle as belonging to the first subgroup when the at least one morphological characteristic satisfies the second set of criteria. In one aspect, the first subgroup of the first group of particles comprises a first type of bacteria and the second subgroup of the first group of particles comprises a second type of bacteria different than the first type of bacteria.

[0015] In one aspect, outputting a result of the classifying of the particle as belonging to the first group of particles comprises generating an overlay visualization indicating that the particle belongs to the first group of particles, and displaying the overlay visualization and the image on a display. In one aspect, the image is a fluorescence image. In one aspect, the method further comprises quantifying a number of particles classified as belonging to the first group of particles, and outputting the number of particles classified as belonging to the first group of particles. In one aspect, the method further comprises inputting a sample comprising the particle into a microfluidic passage of a microfluidic device, immobilizing the particle onto a surface of an electrode disposed in the microfluidic passage with at least one dielectrophoretic force, and capturing the image while the particle is immobilized on the surface of the electrode.

[0016] In some embodiments, a system for classifying a particle in an image as belonging to a first group of particles, the particle being represented in the image as a plurality of contiguous pixels of the image, is provided. The system comprises a microfluidic passage for receiving a sample, the sample comprising the particle, at least one electrode disposed in the microfluidic passage, the at least one electrode configured to immobilize, when activated, the particle onto a surface of the at least one electrode using dielectrophoresis, an optical system configured to capture the image while the particle is immobilized on the surface of the at least one electrode, and at least one computing device. The at least one computing device is configured to determine, based on the image obtained by the optical device, whether an intensity value for the particle in the image is greater than a first threshold value, determine whether at least one morphological characteristic of the particle in the image satisfies a set of criteria associated with particles in the first group of particles, classify the particle as belonging to the first group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image satisfies the set of criteria, and output a result of the classifying of the particle as belonging to the first group of particles.

[0017] In one aspect, the morphological characteristic comprises a size and/or shape of the particle. In one aspect, the size of the particle comprises a number of contiguous pixels representing the particle in the image. In one aspect, the size of the particle is determined based on at least one measurement selected from the group consisting of an area, a circumference, a diameter, an extent, and a bisector. In one aspect, the shape of the particle is determined based on at least one measurement selected from the group consisting of an aspect ratio, an elongation value, a convexity value, a shape factor, and a sphericity value. In one aspect, the particle comprises one of a microorganism or debris. In one aspect, the microorganism comprises one of bacteria, yeast, or mold. In one aspect, the first group of particles comprise microorganisms. [0018] In one aspect, the image includes particles belonging to the first group of particles and particles belonging to a second group of particles, and the at least one computing device being further configured to classify the particle as belonging to the second group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image does not satisfy the first set of criteria. In one aspect, the second group of particles comprises debris. In one aspect, the second group of particles comprises components other than microorganisms.

[0019] In one aspect, when the particle is classified as belonging to the first group of particles, the at least one computing device is further configured to classify the particle as belonging to a first subgroup or a second subgroup of the first group of particles at least in part by determining whether the value for the at least one morphological characteristic satisfies a second set of criteria, and classifying the particle as belonging to the first subgroup when the at least one morphological characteristic satisfies the second set of criteria. In one aspect, the first subgroup of the first group of particles comprises a first type of bacteria and the second subgroup of the first group of particles comprises a second type of bacteria different than the first type of bacteria.

[0020] In one aspect, outputting a result of the classifying of the particle as belonging to the first group of particles comprises generating an overlay visualization indicating that the particle belongs to the first group of particles, and displaying the overlay visualization and the image on a display. In one aspect, the image is a fluorescence image. In one aspect, the at least one computing device is further configured to quantify a number of particles classified as belonging to the first group of particles, and output the number of particles classified as belonging to the first group of particles.

[0021] In some embodiments a method for tuning detector parameters of an automated detector used to automatically detect microorganisms in images is provided. The method comprises processing a plurality of first images with the automated detector to detect a first set of objects in each of the plurality of first images, each first set of objects including particles of interest and other objects, determining detector performance data of the automated detector based on the first set of objects for each of the plurality of first images and a corresponding second set of objects determined from manual identification of objects in each of the plurality of first images, the detector performance data including information about different types of errors generated by the automated detector, tuning the detector parameters of the automated detector based, at least in part, on the detector performance data, processing a second image with the automated detector having tuned detector parameters to detect a third set of objects in the second image, and outputting information about the detected objects in the second image.

[0022] In one aspect, the third set of obj ects includes the particles of interest and other obj ects, and the method further comprises quantifying an amount of particles of interest in the third set of objects. In one aspect, determining detector performance data of the automated detector based on the first set of objects for each of the plurality of first images and a corresponding second set of objects determined from manual identification of objects in each of the plurality of first images comprises comparing object coordinates of objects in the first set of objects and objects in the second set of objects to identify objects detected in both the first and second sets for each of the plurality of first images. In one aspect, an object is identified to be detected in both the first and second sets when a distance between the object coordinates for the object in the first set and the object coordinates for the object in the second set are less than a threshold distance apart. In one aspect, determining detector performance data further comprises computing a pairwise distance matrix between the object coordinates in the first set of objects and object coordinates in the second set of objects, and identifying, based on the computed pairwise distance matrix, objects with a matching set of object coordinates. In one aspect, identifying objects with a matching set of coordinates comprises using a Hungarian algorithm to identify the objects with a matching set of coordinates. In one aspect, tuning the detector parameters of the automated detector is further based, at least in part, on user input assigning weights assigned to each of the different types of errors generated by the automated detector.

[0023] In some embodiments, a method for determining a quantity of microorganisms in a set of image tiles is provided. The method comprises receiving a set of image tiles, the set including a first image tile and a second image tile having an overlap region with the first image tile, detecting, with an automated detector, a first set of microorganisms within the first image tile and a second set of microorganisms within the second image tile, wherein each of the microorganisms in the first set is associated with first object coordinates and each of the of microorganisms in the second set is associated with second object coordinates, identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region, determining the quantity of microorganisms in the set of image tiles based on a number of microorganisms in the first set of microorganisms, a number of microorganisms in the second set of microorganisms and a number of object coordinate pairs identified in the overlap region, and outputting the quantify of microorganisms in the set of image tiles.

[0024] In one aspect, identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region comprises identifying an object coordinate pair when the first object coordinates and the second object coordinates are less than a threshold distance apart. In one aspect, identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region comprises transforming the first object coordinates into a common coordinate space with the second object coordinates, and identifying an object coordinate pair based on the transformed first object coordinates and the second object coordinates. In one aspect, identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region comprises using a Hungarian algorithm to identify the object coordinate pairs.

[0025] In some embodiments, a method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components is provided. The method comprises identifying, based on an image of the sample and a first set of criteria including a first intensity threshold and one or more first morphological criteria, a first set of objects in the sample that have an intensity value in the image above the first intensity threshold and that have a shape characteristic and/or a size characteristic in the image that satisfies the one or more first morphological criteria, the objects in the first set of objects being classified as belonging to a first group of particles associated with the microorganisms of the sample, identifying, based on the image of the sample and a second set of criteria including a second threshold intensity and/or one or more second morphological criteria, a second set of objects in the sample that have an intensity value in the image above the intensity threshold and/or that have a shape characteristic and/or a size characteristic in the image that satisfies the one or more second morphological criteria, the objects in the second set of objects being classified as belonging to the first group of particles associated with the microorganisms of the sample, determining, as a first count, a number of objects in the first set of objects that belong to the first group of particles, determining, as a second count, a number of objects in the second set of objects that belong to the first group of particles, determining the quantity of microorganisms in the sample based on a combination of the first count and the second count, and outputting the determined quantity of microorganisms in the sample.

[0026] In one aspect, determining the quantity of microorganisms in the sample based on a combination of the first count and the second count comprises determining an average of the first count and the second count. In one aspect, determining an average of the first count and the second count comprises determining a weighted average of the first count and the second count. In one aspect, the first set of criteria is selected to minimize false positive detection errors.

[0027] In some embodiments, a method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components is provided. The method comprises identifying, based on an image of the sample and a first set of criteria including a first intensity threshold and one or more first morphological criteria, a first set of objects in the sample that have an intensity value in the image above the first intensity threshold and that have a shape characteristic and/or a size characteristic in the image that satisfies the one or more first morphological criteria, the objects in the first set of objects being classified as belonging to a first group of particles associated with the microorganisms of the sample, providing, as input to a trained statistical model, the image of the sample, wherein the trained statistical model was trained to identify microorganisms in an image based on a plurality of labeled images of samples including microorganisms and other components, determining the quantity of microorganisms in the sample by determining, as a first count, a number of objects in the first set of objects that belong to the first group of particles, determining, as a second count, a number of labeled objects identified in a labeled image output from the trained statistical model, and determining the quantity of microorganisms in the sample based on a combination of the first count and the second count.

[0028] In one aspect, determining the quantity of microorganisms in the sample based on a combination of the first count and the second count comprises determining an average of the first count and the second count. In one aspect, determining an average of the first count and the second count comprises determining a weighted average of the first count and the second count. In one aspect, the first set of criteria is selected to minimize false positive detection errors.

[0029] In some embodiments, a method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components is provided. The method comprises providing, as input to a trained statistical model, an image of the sample, wherein the trained statistical model was trained to identify microorganisms in an image based on a plurality of labeled images of samples including microorganisms and other components, determining based on a labeled image output from the trained statistical model, the quantity of the microorganisms in the sample, and outputting the determined quantity of microorganisms in the sample.

[0030] In one aspect, the trained statistical model was trained to identify a first type of microorganisms in an image, the image provided as input to the trained statistical model is an image of a sample including a second type of microorganisms, and determining the quantity of microorganisms in the sample comprises determining the quantity of the second type of microorganisms in the sample. In one aspect, the image is a fluorescent image. In one aspect, the image is a multi-channel image. In one aspect, the multi-channel image includes a bright field channel and two fluorescent channels.

[0031] In one aspect, the trained statistical model includes at least one trained neural network. In one aspect, the microorganisms are Aspergillus spores. In one aspect, the microorganisms in the sample include a first type of microorganisms and a second type of microorganisms, the image provided as input to the trained statistical model includes both the first type of microorganisms and the second type of microorganisms, and the trained statistical model is trained to label in the output image, only the first type of microorganisms and not the second type of microorganisms. In one aspect, the first type of microorganisms have a different morphology than the first type of microorganisms.

[0032] In some embodiments, a system for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components is provided. The system comprises a microfluidic passage for receiving the sample, at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, particles of the sample on a surface of the at least one electrode using dielectrophoresis, an optical system configured to capture at least one image of the surface of the at least one electrode, and at least one computing device. The at least one computing device is configured to provide, as input to a trained statistical model, an image of the sample, wherein the trained statistical model was trained to identify microorganisms in an image based on a plurality of labeled images of samples including microorganisms and other components, determine based on a labeled image output from the trained statistical model, the quantity of the microorganisms in the sample, and output the determined quantity of microorganisms in the sample.

[0033] In one aspect, the trained statistical model was trained to identify a first type of microorganisms in an image, the image provided as input to the trained statistical model is an image of a sample including a second type of microorganisms, and determining the quantity of microorganisms in the sample comprises determining the quantity of the second type of microorganisms in the sample. In one aspect, the image is a fluorescent image. In one aspect, the image is a multi-channel image. In one aspect, the multi-channel image includes a bright field channel and two fluorescent channels.

[0034] In one aspect, the trained statistical model includes at least one trained neural network. In one aspect, the microorganisms are Aspergillus spores. In one aspect, the microorganisms in the sample include a first type of microorganisms and a second type of microorganisms, the image provided as input to the trained statistical model includes both the first type of microorganisms and the second type of microorganisms, and the trained statistical model is trained to label in the output image, only the first type of microorganisms and not the second type of microorganisms. In one aspect, the first type of microorganisms have a different morphology than the first type of microorganisms.

[0035] In some embodiments, a method for improved detection and/or quantification of microorganisms in a sample, the sample including microorganisms and other components is provided. The method comprises passing a first fluid comprising a stain through a microfluidic passage of a microfluidic device, the microfluidic passage including at least one electrode disposed therein, passing a second fluid through the microfluidic passage after passing the first fluid through the microfluidic passage, the second fluid comprising a controlled solution without the stain, capturing, with an optical system, a first image including a surface of the at least one electrode after passing the second fluid through the microfluidic passage, passing the sample through the microfluidic passage after capturing the first image, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the sample is passed through the microfluidic passage, passing the first fluid through the microfluidic passage while the microorganisms remain captured on the surface of the at least one electrode, passing the second fluid through the microfluidic passage after passing the first fluid through the microfluidic passage and while the microorganisms remain captured on the surface of the at least one electrode, capturing, with the optical system, a second image of the surface of the at least one electrode, and detecting and/or quantifying microorganisms in the sample based on the first image and the second image.

[0036] In one aspect, the method further comprises passing the second fluid through the microfluidic passage prior to passing the first fluid through the microfluidic passage prior to capturing the first image. In one aspect, the method further comprises passing the second fluid through the microfluidic passage prior to passing the first fluid through the microfluidic passage and after passing the sample through the microfluidic passage. In one aspect, passing the first fluid, the second fluid, or the sample through the microfluidic passage comprises pumping the first fluid, the second fluid, or the sample through the microfluidic passage.

[0037] In one aspect, the method further comprises applying at least one first voltage to the at least one electrode as the sample is passed through the microfluidic passage, the at least one first voltage having first characteristics, and applying at least one second voltage to the at least one electrode as the second fluid is passed through the microfluidic passage after the sample is passed through the microfluidic passage, the at least one second voltage having second characteristics different than the first characteristics. In one aspect, the first characteristics include a first amplitude and a first frequency for the at least one first voltage, the second characteristics include a second amplitude and a second frequency for the at least one second voltage. In one aspect, the second amplitude is lower than the first amplitude. In one aspect, the second frequency is different than the first frequency.

[0038] In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises subtracting the first image from the second image to generate a third image, and detecting microorganisms in the sample based on the third image. In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises subtracting the first image from the second image to generate a third image, and quantifying microorganisms in the sample based on the third image. In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises identifying in the first image, a first set of objects, identifying in the second image, a second set objects, removing the first set of objects from the second set of objects, and detecting microorganisms in the sample based on the second set of objects after the first set of objects has been removed.

[0039] In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises quantifying in the first image, a number of first objects, quantifying in the second image, a number of second objects, and quantifying microorganisms in the sample based, at least in part, on the number of first objects and the number of second objects. In one aspect, quantifying in the first image, a number of first objects comprises applying a classifier to the first image to detect the first objects. In one aspect, quantifying, in the second image, a number of second objects comprises applying the classifier to the second image to detect the second objects. In one aspect, the stain is configured to selectively interact with the microorganisms in the sample. In one aspect, the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

[0040] In some embodiments, a system for detecting and/or quantifying microorganisms in a sample, the sample including microorganisms and other components is provided. The system comprises a microfluidic passage for receiving the sample, a pump configured to pump a first fluid comprising a stain, a second fluid comprising a controlled solution without the stain, or the sample through the microfluidic passage, at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, particles of the sample on a surface of the at least one electrode using dielectrophoresis, an optical system configured to capture a first image of the surface of the at least one electrode prior to pumping the sample through the microfluidic passage, and a second image of the surface of the at least one electrode after pumping the sample through the microfluidic passage, and at least one computing device. The at least one computing device is configured to detect and/or quantify microorganisms in the sample based on the first image and the second image.

[0041] In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises subtracting the first image from the second image to generate a third image, and detecting microorganisms in the sample based on the third image. In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises subtracting the first image from the second image to generate a third image, and quantifying microorganisms in the sample based on the third image. In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises identifying in the first image, a first set of objects, identifying in the second image, a second set objects, removing the first set of objects from the second set of objects, and detecting microorganisms in the sample based on the second set of objects after the first set of objects has been removed.

[0042] In one aspect, detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises quantifying in the first image, a number of first objects, quantifying in the second image, a number of second objects, and quantifying microorganisms in the sample based, at least in part, on the number of first objects and the number of second objects. In one aspect, quantifying in the first image, a number of first objects comprises applying a classifier to the first image to detect the first objects. In one aspect, quantifying, in the second image, a number of second objects comprises applying the classifier to the second image to detect the second objects. In one aspect, the stain is configured to selectively interact with the microorganisms in the sample. In one aspect, the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

[0043] In some embodiments, a method for improved detection and/or quantification of microorganisms in a sample, the sample including microorganisms and other components, is provided. The method comprises passing a first fluid comprising a first stain through a microfluidic passage of a microfluidic device, the microfluidic passage including at least one electrode disposed therein, passing a second fluid through the microfluidic passage after passing the first fluid through the microfluidic passage, the second fluid comprising a controlled solution without the first stain, passing the sample through the microfluidic passage after passing the second fluid through the microfluidic passage, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the sample is passed through the microfluidic passage, passing a third fluid comprising a second stain through the microfluidic passage while the microorganisms remain captured on the surface of the at least one electrode, passing the second fluid through the microfluidic passage after passing the third fluid through the microfluidic passage and while the microorganisms remain captured on the surface of the at least one electrode, capturing, with an optical system, an image of the surface of the at least one electrode, and detecting and/or quantifying microorganisms in the sample based on the image.

[0044] In one aspect, the method further comprises passing the second fluid through the microfluidic passage prior to passing the first fluid through the microfluidic passage. In one aspect, the method further comprises passing the second fluid through the microfluidic passage prior to passing the third fluid through the microfluidic passage. In one aspect, passing the first fluid, the second fluid, the third fluid, or the sample through the microfluidic passage comprises pumping the first fluid, the second fluid, the third fluid, or the sample through the microfluidic passage.

[0045] In one aspect, the method further comprises applying at least one first voltage to the at least one electrode as the sample is passed through the microfluidic passage, the at least one first voltage having first characteristics, and applying at least one second voltage to the at least one electrode as the third fluid is passed through the microfluidic passage, the at least one second voltage having second characteristics different than the first characteristics. In one aspect, the first characteristics include a first amplitude and a first frequency for the at least one first voltage, the second characteristics include a second amplitude and a second frequency for the at least one second voltage. In one aspect, the second amplitude is lower than the first amplitude. In one aspect, the second frequency is different than the first frequency.

[0046] In one aspect, detecting and/or quantifying microorganisms in the sample based on the image comprises detecting microorganisms in the sample as objects in the image stained with the second stain but not the first stain. In one aspect, detecting and/or quantifying microorganisms in the sample based on the image further comprises quantifying the detected microorganisms in the sample.

[0047] In one aspect, the optical system includes a first filter and a second filter, and capturing an image of the surface of the at least one electrode comprises capturing the image using the first filter and the second filter. In one aspect, the second stain is configured to selectively interact with the microorganisms in the sample. In one aspect, the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

[0048] In some embodiments, a system for detecting and/or quantifying microorganisms in a sample, the sample including microorganisms and other components is provided. The system comprises a microfluidic passage for receiving the sample, a pump configured to pump a first fluid comprising a first stain, a second fluid comprising a controlled solution without the first stain, a third fluid comprising a third stain, or the sample through the microfluidic passage, at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, particles of the sample on a surface of the at least one electrode using dielectrophoresis, an optical system configured to capture an image of the surface of the at least one electrode after pumping the sample through the microfluidic passage, and at least one computing device configured to detect and/or quantify microorganisms in the sample based on the image.

[0049] In one aspect, detecting and/or quantifying microorganisms in the sample based on the image comprises detecting microorganisms in the sample as objects in the image stained with the second stain but not the first stain. In one aspect, detecting and/or quantifying microorganisms in the sample based on the image further comprises quantifying the detected microorganisms in the sample.

[0050] In one aspect, the optical system includes a first filter and a second filter, and wherein capturing an image of the surface of the at least one electrode comprises capturing the image using the first filter and the second filter. In one aspect, the second stain is configured to selectively interact with the microorganisms in the sample. In one aspect, the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

[0051] In some embodiments, a method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the other components including nucleic acids and/or organelles from degraded cells. The method comprises treating the sample with an exonuclease configured to digest the nucleic acids to oligonucleotides less than five bases long to produce a clarified sample, passing the clarified sample through a microfluidic passage having at least one electrode disposed therein, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the clarified sample is passed through the microfluidic passage, capturing, with an optical system, at least one image of the surface of the at least one electrode, and quantifying microorganisms in the sample based on the at least one image.

[0052] In one aspect, the sample is a sample from a bioreactor. In one aspect, capturing the at least one image comprises capturing a first image prior to passing the clarified sample through the microfluidic passage and capturing a second image after passing the clarified sample through the microfluidic passage, and quantifying microorganisms in the sample comprises quantifying microorganisms in the sample based on the first image and the second image. In one aspect, the method further comprises passing a fluid comprising a stain through the microfluidic passage prior to capturing the at least one image.

[0053] In some embodiments, a method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the other components including at least one surfactant is provided. The method comprises incubating the sample with activated carbon to adsorb the at least one surfactant in the sample to produce a clarified sample, passing the clarified sample through a microfluidic passage having at least one electrode disposed therein, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the clarified sample is passed through the microfluidic passage, capturing, with an optical system, at least one image of the surface of the at least one electrode, and quantifying microorganisms in the sample based on the at least one image.

[0054] In one aspect, the sample is a sample from a bioreactor. In one aspect, capturing the at least one image comprises capturing a first image prior to passing the clarified sample through the microfluidic passage and capturing a second image after passing the clarified sample through the microfluidic passage, and quantifying microorganisms in the sample comprises quantifying microorganisms in the sample based on the first image and the second image. In one aspect, the method further comprises passing a fluid comprising a stain through the microfluidic passage prior to capturing the at least one image.

[0055] In some embodiments, a method for identifying whether a sample is sterile is provided. The method comprises passing a first portion of the sample through a microfluidic passage having at least one electrode disposed therein, generating, with the at least one electrode, at least one dielectrophoretic force that acts on the first portion of the sample, determining, based on a response of the first portion of the sample to the at least one dielectrophoretic force that acts on the first portion of the sample, whether the first portion of the sample comprises one or more microorganisms, and when it is determined that the first portion of the sample does not comprise the one or more microorganisms, labeling the sample as sterile.

[0056] In one aspect, the method further comprises providing an indication that the sample is sterile, and releasing the sample from sterility testing. In one aspect, the method further comprises when it is determined that the first portion of the sample does comprise the one or more microorganisms, labeling the sample as not sterile and providing an indication that a second portion of the sample is to be further processed. In one aspect, the one or more microorganisms comprise bacteria, mold, and/or yeast.

[0057] In some embodiments, a system for identifying whether a sample is sterile is provided.

The system comprises a microfluidic passage for receiving a first portion of the sample, at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, a plurality of particles of the first portion of the sample on a surface of the at least one electrode using dielectrophoresis, an optical system configured to capture at least one image of the surface of the at least one electrode, and at least one computing device configured to determine, based on the at least one image, whether the first portion of the sample comprises one or more microorganisms, and when it is determined that the first portion of the sample does not comprise the one or more microorganisms, labeling the sample as sterile.

[0058] In one aspect, the at least one computing device is further configured to provide an indication that the sample is sterile, and release the sample from sterility testing. In one aspect, the at least one computing device is further configured to when it is determined that the first portion of the sample does comprise the one or more microorganisms, label the sample as not sterile and provide an indication that a second portion of the sample is to be further processed. In one aspect, the one or more microorganisms comprise bacteria, mold, and/or yeast.

[0059] In some embodiments, a method for identifying whether a sample is in a condition for release is provided. The method comprises passing a first portion of the sample into a microfluidic passage having at least one electrode disposed therein, generating, with the at least one electrode, at least one dielectrophoretic force that acts on the first portion of the sample, determining a number of microorganisms immobilized on a surface of the at least one electrode subsequent to the generating the at least one dielectrophoretic force, determining, based on the number of microorganisms immobilized on the surface of the at least one electrode, a concentration of microorganisms in the sample, and when it is determined that the concentration of microorganisms in the sample is less than a threshold concentration, labeling the sample as being in the condition for release.

[0060] In some embodiments, a system for identifying whether a sample is in a condition for release is provided. The system comprises a microfluidic passage for receiving a first portion of the sample, at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, a plurality of particles of the first portion of the sample on a surface of the at least one electrode using dielectrophoresis, an optical system configured to capture at least one image of the surface of the at least one electrode, and at least one computing device configured to analyze the at least one image to determine a number of microorganisms immobilized on a surface of the at least one electrode subsequent to the generating the at least one dielectrophoretic force, determine whether the number of microorganisms immobilized on the surface of the at least one electrode is less than a threshold, and when it is determined that the number of microorganisms immobilized on the surface of the at least one electrode is less than the threshold, labeling the sample as being in the condition for release.

BRIEF DESCRIPTION OF DRAWINGS

[0061] Various non-limiting embodiments of the technology will be described with reference to the following figures. It should be appreciated that the figures are not necessarily drawn to scale.

[0062] FIG. 1 schematically illustrates a system for separation, detection, and quantification of particles in a sample, according to some embodiments of the present technology;

[0063] FIG. 2 illustrates a microfluidic system for separation, detection, and quantification of particles in a sample, according to some embodiments of the present technology;

[0064] FIG. 3 illustrates a static system for separation, detection, and quantification of particles in a sample, according to some embodiments of the present technology; [0065] FIG. 4 is a flowchart of a process for classifying particles in an image, according to some embodiments of the present technology;

[0066] FIG. 5A shows a multi-channel image of an electrode system that may be analyzed, according to some embodiments of the present technology;

[0067] FIG. 5B shows a segmented image based on the fluorescent channel of the image in FIG. 5 A, according to some embodiments of the present technology;

[0068] FIGS. 6A-6C show images segmented using different intensity threshold values, according to some embodiments of the present technology

[0069] FIGS. 7A-7E show multi-channel images of an electrode system having a segmentation overlay presented thereon, according to some embodiments of the present technology;

[0070] FIG. 8A shows a fluorescence image captured by an optical system of microfluidic system, according to some embodiments of the present technology;

[0071] FIG. 8B shows the image of FIG. 8A with a rescaled intensity value, according to some embodiments of the present technology;

[0072] FIG. 8C shows fluorescence images in which a plurality of morphological-based filters have been applied, according to some embodiments of the present technology;

[0073] FIG. 9A shows an illustration of a portion of a software tool used to determine measurement information for a particle in an image, according to some embodiments of the present technology;

[0074] FIG. 9B shows a segmented image after application of a size-based filter, according to some embodiments of the present technology;

[0075] FIGS. 10A and 10B, respectively, show histograms of object area and object brightness, according to some embodiments of the present technology;

[0076] FIG. 11 shows a segmented image after application of a size-based filter, according to some embodiments of the present technology;

[0077] FIG. 12 shows the results of applying an intensity-based filter and a morphological- based filter, according to some embodiments of the present technology;

[0078] FIGS. 13A and 13B, respectively, show histograms of object area and object brightness, according to some embodiments of the present technology;

[0079] FIG. 14A shows an image in which all bright objects and cell-shaped objects have been detected using a two-detector techniques, according to some embodiments of the present technology;

[0080] FIG. 14B shows an image corresponding to the image in FIG. 14A after some duplicate cells are removed, according to some embodiments of the present technology;

[0081] FIGS. 15A and 15B schematically show images for comparing automated detection and manual classification results, according to some embodiments of the present technology; [0082] FIG. 15C shows an image of detector performance evaluation results, according to some embodiments of the present technology;

[0083] FIGS. 16A-16D show images after applying different error weighting strategies, according to some embodiments of the present technology;

[0084] FIG. 17 is a flowchart of an automated process for tuning detector parameters, according to some embodiments of the present technology;

[0085] FIG. 18 schematically shows acquisition of a grid of partially-overlapping image tiles, which may be analyzed, according to some embodiments of the present technology;

[0086] FIG. 19 shows an image of a grid of partially-overlapping image tiles in which particles in the overlap regions are identified and compared, according to some embodiments of the present technology;

[0087] FIG. 20 is a flowchart of a process for quantifying particles (e.g., microorganisms) in s labeled image output from a trained statistical model, according to some embodiments of the present technology;

[0088] FIG. 21 A is a multi-channel image of a spiral electrode system that may be analyzed, according to some embodiments of the present technology;

[0089] FIG. 21B is a zoomed in version of the image of FIG. 21 A;

[0090] FIG. 22 is a bar plot showing the stability of the output of using a trained statistical model to perform classification, according to some embodiments of the present technology; [0091] FIG. 23 is a bar plot showing a comparison of manual counting and automated counting using a trained statistical model, according to some embodiments of the present technology;

[0092] FIG. 24 shows a comparison of different mistake types present when an automated counting technique was used, according to some embodiments of the present technology;

[0093] FIG. 25 is a stacked bar chart showing manual checks of automated detection errors, according to some embodiments of the present technology;

[0094] FIG. 26 is a plot showing bacterial capture results plotted in a row to identify a crossover frequency, according to some embodiments of the present technology;

[0095] FIG. 27A-27C show the results of an automated counting model configured to analyze three-channel images, according to some embodiments of the present technology;

[0096] FIGS. 28A-28B show images of using a trained statistical model to detect particles in an image, according to some embodiments of the present technology;

[0097] FIG. 29A schematically shows components of a microfluidic system, according to some embodiments of the present technology;

[0098] FIGS. 29B-29C schematically illustrate acts for detecting and quantifying microorganisms in a sample using a negative scan image, according to some embodiments of the present technology;

[0099] FIGS. 29D-29E schematically illustrate acts for detecting and quantifying microorganisms in a sample using an integrated image, according to some embodiments of the present technology;

[0100] FIG. 30 is a flowchart of a process for detecting and quantifying microorganisms in a sample using a negative scan image, according to some embodiments of the present technology; [0101] FIG. 31 shows images regarding how negative control scan can reduce the probability of false positive errors, according to some embodiments of the present technology; [0102] FIG. 32 is a flowchart of a process for detecting and quantifying microorganisms in a sample using an integrated image, according to some embodiments of the present technology; [0103] FIG. 33 is a flowchart of a process for using exonuclease to provide a clarified sample prior to processing the sample with a microfluidic device, according to some embodiments of the present technology;

[0104] FIGS. 34-34D show test and control images of using or not using exonuclease to provide samples for processing with a microfluidic device, according to some embodiments of the present technology;

[0105] FIG. 35 is a flowchart of a process for performing sterility testing using a microfluidic device, according to some embodiments of the present technology;

[0106] FIG. 36 schematically shows how a microfluidic device according to some embodiments of the present technology may be used to perform sterility testing; and [0107] FIG. 37 is a flowchart of a process for performing bioburden testing, according to some embodiments of the present technology. DETAILED DESCRIPTION

[0108] Aspects of the technology described herein relate to an apparatus and methods for detecting, separating, quantifying, and/or enriching biological organisms (e.g., bacteria) present in a fluid sample. In particular, the technology described herein provides techniques for rapid detection, separation, purification, and/or quantification of microorganisms in a sample using a microfluidic device comprising one or more electrodes configured to generate dielectrophoretic forces that act on the sample.

[0109] Microbial (e.g., bacterial, yeast, viral and fungal) contamination is a serious and growing global threat to human health and economic development, including pharmaceutical manufacturing and immunotherapy manufacturing and treatment. An example of a conventional technique to assess the presence and degree of microbial contamination in a sample is the Plate Counting Method (PCM). PCM typically includes at least four steps. In step 1, a sample to be analyzed is manually placed in each of multiple test tubes, and the sample in each test tube is diluted to a desired concentration using a buffer solution. In step 2, the diluted samples are plated onto petri dishes containing agar media. Petri dishes including dilution media only (i.e., without the sample) are also plated for use as controls for comparison against the plated diluted samples. In step 3, the plurality of plated samples, the dilution media plates, and empty agar plates are cultured for 24 hours to 14 days to enable microbial particles to grow on the media within the petri dishes. In step 4, the number of bacterial colonies on each of the plates cultured in step 3 is determined, for example, using a microscope.

[0110] PCM is routinely used in medical, pharmacological and food industries to identify bacterial contamination. However, PCM is slow, only moderately sensitive, labor intensive and prone to human errors. For instance, the entire PCM process takes 1-14 days, includes several manual steps in which human intervention is needed, and requires a large number of plated samples at different dilutions and controls. There is therefore a need for new technologies that allow for faster, more sensitive and/or more reliable assessment of microbial contamination in fluid samples.

[0111] Dielectrophoresis (DEP) has shown promise for particle separation; however, it has not yet been applied in clinical settings, pharmaceutical quality control, or manufacturing. For instance, only small sample volumes with unrealistically high bacterial concentrations on the order of 10³-10⁷ CFU/mL have been processed, which limits the applicability of DEP microbial capture methods. DEP particle separation has been achieved only to a limited extent and the separation is restricted to specific cell types, (e.g., separation of Escherichia coli from Bacillus subtilis). Unfortunately, separation of small cells (~lpm in diameter, the size of many bacteria) using DEP has been notoriously difficult. For instance, small bacterial particles undergo significant Brownian motion that adds a time dependent variation in their position, and thus the specificity of separation decreases for small cells, which has previously been thought to limit the applicability of the DEP technique for detecting and/or separating bacteria in a sample.

[0112] Some embodiments of the technology described herein relate to a novel DEP bacterial capture and separation technique (also referred to herein as “Fluid- Screen” or “FS”) that addresses at least some of the limitations of prior DEP techniques.

[0113] Although capture and separation of bacteria from a sample is described herein, it should be appreciated that biological particles other than bacteria, for example, different cells, yeast, mold, fungus, viruses, etc. can also be detected, quantified, separated, and/or enriched using one or more of the techniques described herein. Indeed, the technology described herein has been shown to effectively capture, detect, quantify, and separate a wide range of diverse microorganisms including, but not limited to, both Gram (-) and Gram (+) bacteria, multiple bacterial morphologies, both individual bacteria and cell aggregates, yeasts or molds (including conidia, conidiophores and hyphae), and viruses. Table 1 below illustrates a summary of some microorganisms that have been successfully captured and detected using the techniques described herein.

Table 1: Example microorganisms detected using the techniques described herein [0114] In accordance with some embodiments, a fluid sample containing bacteria is processed in a microfluidic system that includes a microfluidic device. For example, within the microfluidic device, the sample may be subjected to DEP forces to enable separation, detection, enrichment and/or quantification of microorganisms in the fluid sample. Examples of a microfluidic system suitable for use in accordance with the techniques described herein, include the Fluid-Screen Microfluidic System, aspects of which are described in U.S. Patent Application No. 16/093,883 titled “ANALYTE DETECTION METHODS AND APPARATUS USING DIELECTROPHORESIS AND ELECTROOSMOSIS,” filed on October 15, 2018, and U.S. Patent Application No. 14/582,525 titled “APPARATUS FOR PATHOGEN DETECTION” filed on December 24, 2014, each of which is hereby incorporated by reference in its entirety. [0115] FIG. 1 illustrates an example system 100 for detecting bacteria in a sample, in accordance with some embodiments. As shown in FIG. 1, the system 100 comprises a microfluidic device 104 in communication with a computing device 110.

[0116] The microfluidic device 104 may be any suitable device, examples of which are provided herein. In some embodiments, microfluidic device 104 is implemented as a microfluidic chip having one or more passages (e.g., microfluidic channels or chambers) through which a fluid sample 102 is provided for analysis. Although the term “microfluidic channel” or simply “channel” is used herein to describe a passage through which fluid flows through microfluidic device 104, it should be appreciated that a fluid passage having any suitable dimensions may be used as said channel, and embodiments are not limited in this respect. Microfluidic device 104 may comprise a single channel or multiple channels configured to receive a single sample 102 (e.g., to perform different analyses on the sample) or multiple channels configured to receive different samples for analysis. In embodiments having multiple channels, the microfluidic device may be configured to process the single sample or multiple samples in parallel (e.g., at the same or substantially the same time).

[0117] As described herein, sample 102 may include any fluid containing bacteria or other microorganism of interest. In some embodiments, the sample comprises a biological fluid such as saliva, urine, blood, water, any other fluid such as an environmental sample or potentially contaminated fluid, protein matrices, mammalian cell culture, immunotherapy drug product, cell and gene therapy drug product, cell and gene therapy drug sample, bacterial culture, growth media, active pharmaceutical ingredients, enzyme products, or substances used in biomanufacturing, etc.

[0118] As shown, microfluidic device 104 includes at least one electrode 106. The at least one electrode 106 may be configured to receive one or more voltages to generate positive and/or negative dielectrophoresis (DEP) force(s) that act on a sample arranged proximate to the at least one electrode. In some embodiments, the at least one electrode 106 may be configured to receive one or more voltages (e.g., one or more AC voltages) to generate at least one dielectrophoresis force that acts on the sample. The at least one DEP force may cause certain components of the sample to move relative to (e.g., be attracted to or repulsed from) a surface of the at least one electrode 106. For example, in the absence of an electric field, bacteria and other components of the sample 102 may move freely relative to the surface of the electrode. In the presence of the electric field, at least some components (e.g., bacteria) in the sample may be attracted to the electrode surface.

[0119] The small size of bacteria presents an obstacle to optical observation and quantification of bacteria in the sample. The inventors have recognized that activation of the at least one electrode 106 results in an electric field that may be used to selectively trap bacteria on the surface of the electrode(s). When used with an optical system, capturing bacteria on the surface of the electrode(s) may prevent the bacteria from moving in and out of focus of the optical system to enable real-time bacteria detection and quantification, a process referred to herein as “on-chip quantification.”

[0120] The electric field used to capture the bacteria concentrates the bacteria, which enables imaging with fluorescence microcopy or another optical detection technique. Accordingly, bacterial capture using the techniques described herein allows for detection and quantification of bacteria at significantly lower limits compared to some conventional methods, such as the PCM technique described above. The ability to detect and/or quantify bacteria in a sample, even in small amounts, may be useful in applications including, but not limited to, biomanufacturing, gene therapy, analysis of patient samples, vaccine development and/or biothreat detection, antibiotic susceptibility testing.

[0121] For example, the at least one DEP force acting on the sample may cause bacteria (or certain bacteria) to separate from other components of the sample (e.g., via positive DEP). Bacteria in the sample may be attracted to the surface of the at least one electrode 106 allowing for enhanced detection and/or quantification, despite the small size and/or small amount of the bacteria in the sample. Although, microfluidic device 104 is illustrated as having a single electrode, it should be understood that in some embodiments, microfluidic device 104 comprises multiple electrodes arranged in any suitable configuration.

[0122] System 100 may further comprise a computing device 110 configured to control one or more aspects of microfluidic device 104. For example, computing device 110 may be configured to direct the sample 102 into a channel of the microfluidic device. In some embodiments, computing device 110 is configured to control the at least one electrode 106 to generate the at least one DEP force acting on the sample 102. In some embodiments, computing device 110 may cause one or more components of an optical system (not shown) to perform one or more of detection or quantification of the bacteria or other microorganisms in the sample. Non-limiting examples of a computing device 110 that may be used in accordance with some embodiments are further described herein.

[0123] FIG. 2 illustrates an example microfluidic system 200 for detecting the presence of microorganisms (e.g., bacteria) in a sample, in accordance with some embodiments. System 200 includes microfluidic device 208 (e.g., indicated in FIG. 2 as a microfluidic chip) that includes one or more electrodes for generating DEP forces that act on a sample 204 provided as input to the system. Sample 204 may contain microorganisms for which separation, detection, enrichment, and/or quantification may be performed. The one or more electrodes may be arranged in any suitable configuration within the microfluidic device 208. For instance, in embodiments that include multiple electrodes, the electrodes may be arranged in one-dimension along the flow direction of the fluid, perpendicular to fluid flow direction or on a diagonal relative to the fluid flow direction. In some embodiments, a multidimensional (e.g., 2- dimensional, 3 -dimensional) array of electrodes may be used. For instance, a dense array of electrodes arranged both along the direction of fluid flow and perpendicular to the direction of fluid flow may be used.

[0124] As shown in FIG. 2, a flow system 202 is provided. The flow system 202 may provide a solution for transporting the sample 204 to the microfluidic device 208. A first pump 206 may be used to pump the solution and the sample 204 to the microfluidic device 208 at a predetermined flow rate. First pump 206 may be of any suitable type. In some embodiments, first pump 206 is omitted and sample 204 is manually loaded (e.g., using a pipette or capillary flow) as input to one or more channels of microfluidic device 208.

[0125] Microfluidic device 208 is configured to receive sample 204 for processing. Microfluidic device 208 may include one or more passages through which the sample 204 flows. The one or more passages may include at least one electrode formed therein or adjacent thereto. For instance, the at least one electrode may be formed within a passage. The at least one electrode, when activated, is configured to generate an electric field that acts on the sample 204 as it flows through the one or more passages. An electrical system 212 (e.g., a signal generator or controller) is configured to provide one or more voltages to the at least one electrode of the microfluidic device 208 to tune the properties of the electric field for capture of a particular microorganism or microorganisms of interest. Further aspects of the electrical system 212, including example protocols for operating the microfluidic device 208 are provided herein.

[0126] Microfluidic system 200 may include an optical system 210 to facilitate analysis of the sample 204 by performing on-chip quantification. For example, optical system 210 may comprise one or more optical sensors (e.g., a red-green-blue camera) for viewing and/or imaging the sample. The optical sensor(s) may provide for enhanced detection and/or quantification of the captured microorganisms and/or the other components of the sample 204 relative to detection and quantification techniques that require separate culturing of captured microorganisms or an effluent sample from the device. Any suitable optical sensor(s) may be used. In some embodiments, the optical sensor(s) comprises a digital camera. In some embodiments, the optical sensor(s) comprises electronic sensors including CMOS compatible technology. In some embodiments, the optical sensor(s) comprise fiber optics. However, any suitable optical sensor(s) may be used. In some embodiments, bacteria in the sample are stained

(e.g., with a fluorescent dye) and the optical system 210 is configured to perform microscopy

(e.g., fluorescence microscopy) of captured stained bacteria. In some embodiments, optical system 210 is configured to capture one or more images (e.g., color images) of the at least one electrode while the sample is flowing through the microfluidic device 208. In some embodiments, the detector comprises nanowire and/or nanoribbon sensors. In some embodiments, the field of view of the optical system 210 at a particular magnification is insufficient to capture the entire surface of the one or more electrodes. In such embodiments, the optical system 210 may be configured to capture multiple partially overlapping images that collectively cover the entire surface of the one or more electrodes. The multiple captured images may then be analyzed to detect and/or quantify the microorganisms of interest in the sample.

[0127] Microfluidic system 200 also includes computer 230 configured to control an operation of optical system 210 and/or to receive images from optical system 210 and to perform processing on the received images (e.g., to count a number of microorganisms trapped by the microfluidic device 208). In some embodiments, the received images are analyzed to determine the number of microorganisms captured by the at least one electrode. For instance, microorganisms may be identified in the received images as spots (e.g., fluorescent spots) located on the edges of the electrodes. In this way, a captured target microorganism species may be differentiated from other components in the sample that are not captured and may appear as floating above the at least one electrode or located between electrodes.

[0128] After the sample 204 is processed by the microfluidic device 208 and/or optical system 210 to capture and/or quantify microorganism immobilized on the electrode(s), the sample 204 may be removed from the microfluidic device 208. For example, a second pump 216 may be provided for pumping the sample 204 out of the microfluidic device 208. The second pump 216 may be of any suitable type. In some embodiments, microfluidic system 200 comprises a flow sensor 214 for measuring a flow rate at which the sample 204 is removed from the microfluidic device 208. The flow sensor 214 and the second pump 216 may be in communication to control a flow rate at which the sample 204 is removed from the microfluidic device 208.

[0129] As described herein, microfluidic system 200 may be used for separating bacteria from other components (or for separating certain bacteria from other bacteria) in sample 204. Microfluidic system 200 comprises a waste region 218 arranged to receive other components of the sample 204 which have been separated from the bacteria by the microfluidic device 208 and subsequently removed from the sample 204, for example, using the second pump 216. In the description below, analysis of the fluid collected in waste region 218 may be referred to as analysis of the “effluent sample.” Microfluidic system 200 may further include effluent region 220 for receiving a purified version of sample 204 containing substantially only target microorganisms that were captured using microfluidic device 208.

[0130] In some embodiments, an amount of time needed to process a sample using microfluidic system 200 is substantially less than an amount of time required to process a sample using a conventional PCM sample processing system. As shown in FIG. 2, processing a sample using system 200 may include at least three steps. In step 250, a sample is provided as input to microfluidic system 208 and bacteria are captured from the sample in the presence of an applied electric field. In step 260, automated on-chip quantification is performed, for example, using computer 230 to analyze one or more images recorded by optical system 210. In step 270, further analysis may be performed on waste 218 and/or effluent sample 220, as desired. In sum, the entire process for detecting and/or quantifying microorganisms in a sample using system 200 may take on the order of minutes or an hour to a few hours, which is substantially faster than the multiple days (e.g., 1 to 14 days) typically required to process samples using PCM. [0131] In some embodiments, rather than pumping sample 204 through one or more passages through which the sample flows, sample 204 may be manually provided as input to microfluidic device 208 for analysis. For instance, one or more droplets of sample 204 may be provided as input to microfluidic device 208 using a pipette or other suitable technique. In such embodiments, the sample is analyzed in a “static” condition rather than in a condition in which microorganisms are captured by the at least one electrode as the sample flows past the electrode(s) (e.g., as in the case of microfluidic system 200 as shown in FIG. 2). FIG. 3 illustrates a microfluidic system 300 for detecting microorganisms in a sample, according to some embodiments. As shown, microfluidic system 300 may include many of the same components as microfluidic system 200, but may omit certain components of the microfluidic system 200, such as the first pump 206, which are not needed when the sample is manually provided as input to the microfluidic device 308 (indicated in FIG. 3 as a static microfluidic chip).

[0132] Existing techniques for detecting microorganisms in fluid samples (e.g., water and other fluids) may be inefficient in several ways including, but not limited to, their inability to detect low levels of contaminant and/or their inability to culture certain types of microorganisms. In some instances, existing detection methods may take days to provide results. While faster methods such as quantitative polymerase chain reaction (qPCR) can reduce the response time to a few hours, such methods require complex sample preparation, high costs, have limited portability, and cannot be used for process streamlining.

[0133] Some embodiments relate to computational methods for analyzing one or more images captured by an optical system of a microfluidic system including microfluidic device configured to capture particles of interest using dielectrophoresis, as described above. FIG. 4 illustrates a process 400 for classifying a particle in an image as belonging to a first group of particles or a second group of particles. In act 410, and image of a surface of one or more electrodes of a microfluidic system (e.g., microfluidic systems as shown in FIGS. 2 and 3) is received. For instance, as described above, a microfluidic device of the microfluidic system may be controlled to generate a positive dielectrophoretic force that results in one or more types of particles (e.g., bacteria) to be attracted to the surface of the electrode(s). While the particles are attracted to the electrode surface, one or more images of the electrode surface may captured by an optical system of the microfluidic system. Such an image may be received in act 410 by a computing device (e.g., computing device 110, computer 230, computer 330) for analysis. [0134] FIG. 5A provides an example of such a type of image. In the example of FIG. 5 A, a multi-channel image (one fluorescence channel and one bright-field channel) of electrodes on which Bacillus megaterium cells have been captured using a microfluidic device is shown. In the example image of FIG. 5A, the B. megaterium cells were stained with the fluorescent stain sybr green, and imaged at 20x magnification using an Olympus BX-63 microscope. However, it should be appreciated that the techniques described here are not limited to processing images captured using any particular type of optical system.

[0135] After receiving the image in act 410, process 400 proceeds to act 412, where it is determined whether there are more particles in the image to classify. If it is determined in act 412 that there are additional particles to classify, process 400 proceeds to act 414, where it is determined whether an intensity value associated with a particle to be classified is above a threshold value. If it is determined in act 414 that the intensity value associated with the particle is not greater than the threshold value, process 400 returns to act 412, where it is determined whether there are more particles to classify. In some embodiments, rather than comparing the intensity value for each particle in an image with a threshold value, the entire image may be segmented based on intensity such that all pixels having an intensity less than the threshold value are disregarded. Intensity -based image segmentation in accordance with some embodiments is described in more detail below.

[0136] If it is determined in act 414 that the intensity value associated with the particle is greater than the threshold value (or alternatively, if intensity-based segmentation has been applied), process 400 proceeds to act 416, where it is determined whether at least one morphological characteristic associated with the particle satisfies a first set of criteria (e.g., size and/or shape criteria). If it is determined in act 416 that the at least one morphological characteristic of the particle does not satisfy the first set of criteria, process 400 proceeds to act

418, where the particle is classified as belonging to a second group of particles (e.g., debris).

Process 400 then returns to act 412, where it is determined whether there are more particles in the image to classify and the process repeats. If it is determined in act 416 that the at least one morphological characteristic of the particle satisfies the first set of criteria, process 400 proceeds to act 420, where the particle is classified as belonging to the first group of particles (e.g., cells).

Process 400 then proceeds to act 422 where it is determined whether the particle classified as belonging to the first group of particles is to be classified into a subgroup (e.g., certain types of cells). If it is determined in act 422 that sub-group classification is not desired, process 400 returns to act 412, where it is determined whether there are more particles in the image to classify and the process repeats. If it is determined in act 422 that sub-group classification is desired, process 400 proceeds to act 424, where it is determined whether one or more morphological characteristics of the particle satisfy a second set of criteria (e.g., size and/or shape criteria) associated with a sub-group of the first group of particles. Process 400 then returns to act 412, where it is determined whether there are more particles in the image to classify and the process repeats until it is determined in act 412 that there are no further particles in the image to classify, and process 400 proceeds to act 426, where information determined based on the particle classification is output. For instance, all particles identified as belonging to the first group of particles may be identified and/or quantified, and an output of the identification and/or the quantification may be provided to a user, examples of which are described below.

[0137] The cell-specific detection and cell-debris classification techniques described herein include data from an experiment in which Bacillus megaterium cells were captured on electrodes of a microfluidic device (e.g., the microfluidic device shown in FIG. 2 or 3), stained with the fluorescent stain sybr green, and imaged at 20x magnification using an Olympus BX-63 microscope as the optical system. However, it should be appreciated that the techniques are not limited for use with any particular type of image capture setup. In a first “cell-specific” detection technique described below, it was demonstrated how a data set containing both cells and debris (from a nonspecific detection) can be reduced to contain only data on cells by restricting the image segmentation parameters and filtering out remaining unwanted objects. In a second “cell- debris classification” technique described below, the properties of both cells and debris are classified by performing a nonspecific detection followed by classification based on the distinctive morphological characteristics of the cells and/or the debris.

Cell-specific detection

[0138] In some embodiments, it is desirable to process an image using intensity-based segmentation and morphological filters to identify particles of interest (e.g., cells) while excluding all other types of objects (e.g., debris, pixel noise, electrode features, etc.) represented in the image. Image segmentation is the process of separating regions of pixels based on their intensity values. For instance, FIG. 5B shows the fluorescence channel of the multi-channel image of FIG. 5 A following segmentation. In some embodiments, segmentation is performed using grayscale images, fluorescence images, or other color images (using each color channel (e.g., red, green, blue) or alternative representations like hue, saturation, value). The minimum intensity value used to differentiate objects from the background may be a useful parameter in detecting only the desired objects (e.g., captured particles) during image segmentation, while excluding all other objects.

[0139] Intensity values from an 8-bit camera range from 0 to 255. As shown in FIG. 6A, by performing a very permissive segmentation with a minimum intensity threshold value of 2, all particles of interest (e.g., bacteria) may be detected as well as some debris (e.g., dust), pixel noise, and other objects that are not of interest (e.g., microfluidic device features). As shown in FIG. 6B, by increasing the minimum intensity threshold value from 2 to 5, much of the pixel noise is removed, but the particles of interest and some of the debris remains. In the present example, because the particles of interest are substantially brighter compared to the debris, nearly all of the debris and microfluidic device features may be removed by setting a much higher intensity threshold value for segmentation. For example, the segmented image shown in FIG. 6C was generated by setting the intensity threshold value for segmentation at a value of 75. By performing segmentation that restricts detection to very bright objects, much of the debris in the image can be excluded. In the example images shown in FIGS. 7A-7E, the original fluorescence image is presented in green as indicated, while objects detected by particle-specific detection using intensity -based segmentation have a pink overlay as indicated. Similar to FIGS. 5A-5B, the images shown in FIGS. 7A-7E are multi-channel images that include a fluorescence channel and a bright field channel to visualize the electrodes in the image.

[0140] As should be appreciated from the foregoing, intensity-based segmentation is applied to an image in some embodiments to include particles of interest (e.g., cells) and exclude debris and other objects that are not of interest. However, such an approach requires the particles of interest to be brighter (i.e., have higher intensity values) in the image than the debris. In some embodiments, a series of acts may be performed to determine whether intensity -based segmentation can be used to distinguish particles of interest and debris in a fluorescence image. The series of acts may include:

• Rescale the image intensity (contrast setting) such that all fluorescent objects in the image are clearly visible to a user. For these images, rescaling the intensity such that all pixels with an intensity value > 30 are displayed at maximum brightness may reveal even dim fluorescent regions in the original image. By rescaling in this manner, the dynamic range of the image is reduced, such that the ability to distinguish between medium and high brightness objects as well as the internal structure of bright objects may not be possible. In the examples shown in FIG. 8B, it can be seen that objects invisible in the native contrast setting (full dynamic range) of the original image in FIG. 8A become visible when the intensity is rescaled.

• Step through varied intensity cutoffs to observe which objects are included or excluded. This act may be achieved using software that provides a real-time overlay of objects within the specified intensity range as the intensity range is adjusted with sliders or by inputting new values. In this way, a lower intensity bound for detected objects can be established. By viewing which regions are above or below a given threshold, it can be evaluated whether an intensity threshold value exists that enables distinguishing particles of interest (e.g., cells) from debris on the basis of their intensity values. In the image shown in FIG. 8A, it was found that much of the debris was very dim (intensity value ~10) compared to the cells (intensity value >75), so it was possible to select an intensity threshold value that provided a substantial reduction in detected objects simply by setting the threshold value high enough to discard most of the debris, but low enough to include all of the cells.

[0141] In some embodiments, intensity -based image segmentation may not be sufficient to distinguish particles of interest and other objects (e.g., debris) in the image. Accordingly, in addition to using intensity-based image segmentation, morphological features of the particles of interest (and/or of the objects to be ignored) may be used to classify particles in an image. For instance, as discussed above in connection with FIG. 4, at least one morphological characteristic of the particle to be classified in the image may be compared against a set of criteria (e.g., size and/or shape criteria) to determine whether to classify a particle as belonging to a first group of particles (e.g., cells). In some embodiments, classification based on morphological characteristics of particles in the image may be implemented using one or more filters that describe distinctive morphological characteristics of the particles of interest. In the example shown in FIG. 8C, a plurality of morphological-based filters were applied to distinguish microbes of interest from unwanted objects (e.g., debris) in the image. In particular, in the example of FIG. 8C, only objects with an area 1 -400pm², mean intensity >80, and shape factor <

0.75 were classified as cells and all other objects were considered to be debris. Morphological- based filters may be defined in any suitable way depending on the morphology of the particles of interest. In some embodiments, automated techniques including, but not limited to, applying a variance maximization technique (like principal component analysis or singular value decomposition) to the space of object measurements followed by a clustering technique (e.g. k- means clustering) is used to define suitable filters.

[0142] As described above, appropriate filters used to remove unwanted objects in accordance with some embodiments, may be defined based on one or more morphological characteristics of the particles of interest in the image and/or the unwanted objects (e.g., debris) to be excluded. Particles of interest that have distinctive morphological characteristics in terms of size and/or shape, when compared to the debris in the image, are better candidates for defining morphological-based filters to distinguish the particles of interest from debris automatically with little or no manual inspection required. In some embodiments, to develop an understanding of how particles of interest (e.g., cells) can be distinguished from debris remaining after intensity -based image segmentation, the following steps are performed:

• Choose a set of measurements to be performed on every detected object. For example, a set of measurements describing each object’s size (e.g., area, circumference, diameter, extent, bisector, ferret) and each object’s shape (e.g., aspect ratio, elongation, convexity, shape factor [circularity], sphericity) may be selected.

• Inspect particles of interest (e.g., cells) in images that are known to be cells to develop a better quantitative understanding of the shape and size of both cells and debris. This understanding may be achieved, at least in part, using software tools that allow for object interrogation and measurement determination us to click on an object in the image and displays its measurements, an example of which is shown in FIG. 9A.

• Identify a suitable set of measurements to distinguish cells from debris. For example, one or more measurements for which cells and debris occupy different ranges of values from one another may be determined. In the above example, two distinct types of debris were identified: very small specks much smaller than a cell and, occasionally, large pieces of dust much larger than a cell. Due to this characterization of the debris, a size-based filter was applied to distinguish cells from debris. Additionally, the small specks of debris were nearly perfectly round, while all the cells had a somewhat elongated sausage-like shape. Due to this characterization of the cells, a shape-based morphological filter was applied. In some embodiments, using a minimal set of measurements that distinguishes particles of interest (e.g., cells) and debris may be used. Applying filters to many measurements and requiring objects to meet all of the morphological criteria in the set of criteria described by the filters to be classified as a particle of interest runs the risk of inadvertently discarding particles of interest due to the criteria in the set of criteria being too strict.

• Identify suitable values for each measurement. At this step, cutoff values for each measurement filter are identified. Ideally, cutoff values that include all cells and exclude all debris would be identified. In some embodiments, the cutoff values may be determined by identifying, for each measurement filter being used, a transition point for the measurement value between cells to debris. In some embodiments, the identification is performed automatically.

• Apply filters and inspect remaining objects. In some embodiments, this step is accomplished with an overlay on the original image that superimposes the detected objects in a distinctive color, subject to the filtering requirements that have been defined, allowing for a visualization of which objects remain following application of the defined morphological filters. FIG. 9B shows an example, in which after application of a size filter, a large piece of debris is shown as being successfully excluded.

[0143] In some embodiments, after the intensity-based image segmentation and morphological filters have been applied, distributions of object properties may be determined to evaluate the effect of applying different amounts of image segmentation and morphological filtering. Table 2 shows the number of objects detected using very general detection of all bright objects to a more specific detection geared towards only detecting cells. The increasing specificity is apparent by the drastically reduced number of objects in each set as more restrictive segmentation/filtering is applied.

Table 2: Quantification of detected objects using different segmentation/morphological filtering techniques [0144] With more restrictive segmentation and morphological filtering, it was observed that the distributions of object measurements shifted from a dataset that was dominated by pixel noise and debris to one that is dominated by cells. FIGS. 10A and 10B illustrate histograms of object area and object brightness for the four data sets represented in Table 2. It can be observed that when including the debris and pixel noise (i.e., “All Objects”), the data is dominated by small, dim objects. By contrast, when restricting detection based on intensity thresholds and applying appropriate morphological filters, more accurate measurements of the characteristic size and brightness of the cells are revealed.

Cell-debris classification

[0145] In the example above, only objects in an image corresponding to particles of interest (e.g., cells) were classified, whereas all other objects were excluded from further analysis. In some embodiments, it may be desirable not to discard measurements of detected debris objects, but rather to characterize those objects separately from the population of particles of interest (e.g., cells). For instance, as discussed with regard to process 400, particles that have intensity values above a threshold value (e.g., that meet the intensity -based segmentation criterion), but that do not satisfy a first set of morphological criteria, may be classified as belonging to a second group of particles (debris), whereas, particles that satisfy the first set of criteria may be classified as belonging to the first group of particles (cells). Such an analysis may be beneficial for a more thorough analysis of the composition of the entire sample, for example, to consider both the particles of interest and other particles in their environment. In such embodiments, both particles of interest and debris may be segmented and separated into classes based on their intensity and/or morphological characteristics, enabling for measurements distribution analysis for each class separately.

[0146] As an example, an intensity-based filter and a morphological filter may be applied to an original image to exclude pixel noise. Pixel noise refers to small groups of pixels slightly above background intensity, but not caused by any physical object of the microfluidic device

(e.g., the electrodes). Rather, pixel noise originates from electrical fluctuations at the level of the optical system’s sensor and is typically present in any image obtained by a digital image sensor

(e.g., a CCD or CMOS image sensor). In one example, a minimum intensity value of 5 was set to detect both cells and debris in an image. Additionally, a size-based morphological filter was imposed to require the object size to be at least 5 pixels (e.g., to exclude pixel noise). The result of applying these two constraints to an original image is shown in FIG. 11, in which 6655 objects were identified. Determining the lower intensity threshold (5 in the example above) was determined using the following steps:

• Rescale the image intensity (e.g., by adjusting a contrast setting of the image) such that all fluorescent objects in the image are visible in the image.

• Identify a threshold value, which detects all bright objects but excludes pixel noise. The threshold value may be set high enough to exclude pixel noise, but low enough to not exclude particles of interest (e.g., cells) or debris in the image.

[0147] After setting the low threshold value, one or more cell-specific filters determined based on knowledge about the particles of interest may be used to classify those particles from debris. For instance, in the example described herein for classifying B. megaterium cells, the following criteria may be used to define a cell-specific filter: object area 10-4000pm² and mean intensity value > 30. In this example, using this classification resulted in 3102 of the 6655 objects (47%) being classified as B. megaterium cells and other objects being classified as debris. As described above, in some embodiments the process of determining classifiers is performed using automated technique designed to identify meaningful axes of separation of object measurements and cluster objects in that space.

[0148] FIG. 12 shows the results of classifying objects identified in FIG. 11. Objects classified as cells are shown in green, whereas objects classified as debris are shown in white. [0149] In this example, rather than discarding (filtering out) objects classified as debris, those objects are labelled as a separate class, which can be characterized. Because object boundaries may depend strongly on the intensity threshold used for segmentation, object measurements can vary between a cell-specific detection with a high threshold and an all-bright- objects detection with a low threshold. An example workflow for cell-specific detection in accordance with some embodiments is as follows:

• Choose a set of measurements to be performed on every detected object.

• Inspect the measurements on a subset of objects that are definitely cells and a subset of objects that are definitely debris.

• Identify a suitable set of measurements to distinguish cells from debris. It should be appreciated that different sets of measurements could lead to the same final classification, given the redundancy between certain measurements of size (e.g. diameter, radius, extent and bisector). In some instances, cells may generally be brighter than debris and may occupy a distinct range of sizes compared to small debris specks and large dust fibers.

• Identify suitable cutoff values for each measurement to define two classes (e.g., cells and debris), as discussed above.

• Create a classifier from the determined cutoff values.

[0150] In some embodiments, after the intensity-based image segmentation and morphological filters have been applied, distributions of object properties may be determined separately for each class, as shown in FIGS. 13A and 13B. In this example, it was observed that that B. megaterium cells exhibit a characteristic size distribution centered around lOOpm² and have a mean intensity value around 60. By contrast, debris objects are characterized as being substantially smaller and dimmer (i.e., have lower intensity values) compared to cells as shown in the plots of FIGS. 13A and 13B, respectively.

[0151] As described above, some embodiments of the present technology relate to novel computational methods to enable real-time detection and quantification of particles of interest (e.g., microbes, cells) captured by one or more electrodes of a microfluidic device (e.g., a microfluidic device described in connection with FIG. 2 or 3). In some embodiments, detector performance is optimized through comparison of the automated results with manually curated data. In the specific example, described below, a pair of object detectors are configured to analyze fluorescence microscope images to identify microbial cells and other visible objects. To evaluate detector performance, the results of these detectors were compared against a manual (i.e., human) evaluation of the same images. Using numerical optimization, detector parameters were tuned to accommodate different applications.

[0152] In some embodiments, raw data produced by a microfluidic system (e.g., the microfluidic system shown in FIG. 2 or 3) after capture, staining, and imaging of microorganisms in a fluid sample is a grid of multiple (e.g., hundreds or thousands) of partially- overlapping images (e.g., fluorescence microscope images) covering the capture area of a microfluidic device within the microfluidic system. To facilitate analysis of the large number of images, some embodiments relate to an automated computer vision technique to identify and enumerate objects visible in the images. In some embodiments, object detection is achieved by identifying contiguous regions of bright pixels in an image. For instance, multiple binary images may be created at different intensity thresholds, and contours (e.g., the boundary between dark and light regions) may be identified. Closed contours of a given size and shape may be identified as potential object, subject to a stability requirement across multiple thresholds.

[0153] The inventors have recognized and appreciated the tuning the parameters of the automated computer generation technique is important to ensure particles of interest (e.g., cells, microbes) in the images are accurately classified. For instance, the series of binary images may be created using a specified minimum, maximum and step size for intensity thresholding. Additionally, intermediate contours used to filter objects based on their shape and/or size, may use minimum and maximum allowed values for area, circularity, inertia and/or convexity. Yet further, the stability of intermediate contours, which may be used to determine the final objects can be specified by a repeatability across thresholds and minimum distance cutoff for the center of the objects. The objects that a given detector finds (or fails to find) may depend, at least in part, on appropriately specifying values for each of these parameters.

[0154] In some embodiments, two separate object detectors are applied to each image. The first detector (referred to herein as the “all objects” detector) is configured to identify all objects in the image above a background intensity value. The second detector (referred to herein as the “all cells” detector) is configured to apply an identical intensity threshold filters as the all objects detector, but is further configured to additional require an additional morphological filter (e.g., shape filter) such that intermediate contours have a specific shape, quantified, for example, by particular ranges of circularity, inertia, and convexity. The particular ranges can be tuned to accommodate the characteristic shape of a given particle of interest (e.g., a particular microbe or cell). The output of the two detector approach is two sets of coordinates (one set for each detector) describing the positions within the image. FIG. 14A illustrates an image in which all bright objects (teal circles, as indicated) and all cell-shaped objects (pink circles, as indicated) in a fluorescence microscope image have been automatically detected using the two detector approach described herein.

[0155] In the output of the two-detector approach, all cells are detected twice (once in the all objects detector and once in the all cells detector). To facilitate quantification of the cells in the image, some embodiments relate to performing duplicate removal to remove doubly -counted cells from the all objects detector, and instead categorize the objects in the image as cell-shaped objects and other objects for subsequent analysis (e.g., quantification), as shown in FIG. 14B.

Because the positions of the detected objects output from each of the detectors are determined, in an ideal case, objects identified by both detectors could be identified as those that have identical positions in the image. However, because the two detectors may be configured to use different parameters (e.g., different intermediate contours) to determine the final set of objects, the object positions across detectors may not match exactly. To account for this position mismatch, some embodiments are configured to implement a best (e.g., close, but not exact) matching between the object positions in the two data sets output from the detectors. The result of applying the best matching technique is two sets of coordinates - a first set of cell-shaped objects detected in the image, in which each cell is only represented once, and a second set of other objects, in which each other object is only represented once.

[0156] In some embodiments, the results output from the automated detection technique described above is automatically compared with manual classification to evaluate the accuracy of the automated dual-detector technique. In this example, manual enumeration and classification were performed using the Cell Counter plugin of the open source ImageJ (available from https://imagej.nih.gov/ij) application. Images were inspected under various brightness and contrast settings to identify bright objects as well as very dim objects, often barely above background intensity. Objects were manually classified as cells or other objects. Object coordinates corresponding to the detected cells and other objects were exported in xml format for subsequent use. Automated classification of the same images was performed using the above-described dual-detector technique, with the output of the object coordinates for detected cells and other objects also being recorded.

[0157] In some embodiments, a comparison of the manually classified objects and the automatically classified objects was performed by identifying matchings that minimize the distance between objects from the two sets of coordinates (automated and manual). In some embodiments, the Hungarian algorithm was used to match objects from automatic detection to those from manual enumeration. FIG. 15A shows that the best matching techniques (e.g., using the Hungarian algorithm) finds the best possible matchings between two sets of coordinates, the objects found by automatic detection, and those labelled by manual classification. The best matching approach enables object matching even when coordinates do not match perfectly because of, for example, a discrepancy between where a human selects an object location and where the automated detector calculates the center of the object.

[0158] In some embodiments, candidate matching objects having a distance above a threshold distance (e.g., a typical length of a cell or other particle of interest) were excluded to address the possibility that the two data sets may have detected different objects due, for example, to a detection error. For example, FIG. 15A shows that the two objects detected in the lower portion of the image are separated by a substantial distance. FIG. 15B shows that in some embodiments, pairings further than a cutoff (threshold) distance (e.g., the two objects detected in the lower portion of the image) are excluded as matching objects. The result of the best matching process may be an accurate quantification of unmatched objects from manual classification (false negatives) and unmatched objects from automated detection (false positives).

[0159] In some embodiments, best matching between two sets of coordinates A and M (output from the annotated and manual detectors, respectively) is performed according to the following steps:

• Compute the pairwise distance matrix Dij between Ai and Mj

• Set Dij > threshold to infinity, prohibiting these from being used as matches

• Use the Hungarian algorithm to find the matches between Ai and Mj that minimize the total distance between pairs, that is, the best possible set of (allowed) matchings

• Return the set of matchings as well as the unmatched objects from each data set [0160] In some embodiments, the automated detector performance may be evaluated by considering a set of objects A produced by the automated dual-detector system described herein comprising automatically detected cells (Ac) and automatically detected other objects (Ao), and a set of objects from manual classification, M, with some labelled as cells (Me) and some labelled as other objects (Mo). Best matching according to the techniques described above may be performed, and each pairing, as well as each unpaired object, may represent a distinct outcome. Objects present in the manual classification but not detected using the automated detection are considered as false negatives, whereas automatically detected objects absent from the manual classification are considered as false positives. The complete set of possible outcomes is:

Complete agreement between manual and automatic:

• True cell (Ac matched with Me)

• True other (Ao matched with Mo)

Manually classified object not detected:

• False negative cell (unmatched Me)

• False negative other (unmatched Mo)

Detected object not manually classified: • False positive cell (unmatched Ac)

• False positive other (unmatched Ao)

Disagreement in classification:

• False classification cell (Ac matched with Mo)

• False classification other (Ao matched with Me)

[0161] FIG. 15C shows a graphical depiction of detector performance evaluation in accordance with some embodiments. To generate the graphical depiction in FIG. 15C, an automated dual-detector system located cells (identified as “AUTO CELLS”) and other objects (identified as “AUTO OTHER”) in an image. Manual classification of cells (identified as “MANUAL CELLS”) and other objects (identified as “MANUAL OTHER”) was also performed. Best matching using the techniques described herein between the automated detection objects and manually classified objects was performed, and pairings represented by lines between circle centers are shown. Unmatched objects (false negatives/positives) are represented by bold circles. In the example shown in FIG. 15C, there were seven correct pairings, two false classifications, and two false positives.

[0162] As described above, some embodiments are configured to automatically tune automated detector parameters based, at least in part, on the automated detector performance determined using the techniques described herein. For instance, a quantity that describes detector performance on a given image may first be determined by, for example, counting each type of mistake represented in the detector performance measure, and weighting each mistake based on how undesirable the mistake is for a given application. For example, for some applications, failing to detect a cell entirely (a false negative cell error) may be a more important error than other types of errors. False negative other errors may be slightly more permissible, since an object annotated as a non-cell is less likely to pose a serious contamination threat. False positive errors may be even more permissible, as they may require further manual inspection of the images even though they may not contain particles of interest. False classifications may represent the least severe type of error, since these objects are still detected and therefore can be flagged for manual review. In some embodiments, a set of weights are assigned to different types of errors, and example of which is shown below in Table 3.

Table 3: Example weights assigned to different error types.

[0163] In some embodiments, the weights assigned to each of the different error types may be used to construct a score. An example of such a score using the weights in Table 3 is as follows: score = 8 * FNc + 4 * FNo + 2 * FPc + 2 * FPo + 1 * FC [0164] where FNc is the number of false negative cell errors, FNo is the number of false negative other errors, FPc is the number of false positive cell errors, FPo is the number of false positive other errors, and FC is the number of false classification errors. In this formulation, a larger score may indicate worse agreement between automated detection and manual classification for a given image (i.e., a high score represents poor detector performance.

[0165] By defining a function (the score) of many variables (the automated detector parameters), numerical optimization is used in some embodiments to determine an optimum set of automated detector parameters by, for example, searching for a global minimum in the data. In some embodiments, such numerical optimization is performed by using simulations to explore various combinations of detector parameters with the goal of minimizing the score for a given set of images (a training data set). Such an approach enables automated identification of detector parameters for any given type of experiment (e.g., a given microorganism in a given media) and a given set of preferences regarding which types of errors are more permissible than others (the weights).

[0166] In some embodiments, numerical optimization is performed using differential evolution, which simulates an evolutionary process where a population of candidate solutions (combinations of automated detector parameters) is initially spread out in the parameter space, evaluated on their fitness (the score), and mutated to incorporate the traits of other candidates. Eventually the candidate solutions converge to the likely global minimum in the function to obtain an optimized set of automated detector parameters for a given application. Provided that the images obtained in subsequent experiments are similar enough to the images in the training data set, manual classification and numerical optimization using differential evolution may need only be performed once to optimize automated detector parameters for each new type of application.

[0167] To examine the score improvement offered by differential evolution compared to manually tuned parameters, manual enumeration as described above was performed on a training data set containing 15 images and 865 objects. Coarse hand tuning of parameter was achieved by using the score along with graphical depictions, an example of which is shown in FIG. 15C. Differential evolution was performed over parameter ranges identified during handtuning and yielded a 25% reduction in total score over the 15 image set compared to the bestperforming set of hand-tuned parameters. Because the score alone did not indicate which type of errors were occurring, the rate of each type of mistake was also calculated. For example, the rate of false negative cells is given by the number of false negative cells compared to the number of manually classified cells in the data set. The rate of false positive cells is given by the number of false positive cells compared to the total number of automatically detected cells in the data set. With these definitions, the total score and mistake distribution over the 15 image set was determined, with the results shown in Table 4. As shown, the improvement in score was largely driven by a reduction in false negatives.

Table 4: Total score and mistake distribution using the weights in Table 3 [0168] As discussed above, a unique strength of the detector parameter optimization technique described herein is the ability of the technique to modulate detector performance to favor certain types of errors by optimizing detector parameters to a differently weighted score. For instance, although false negative cell errors are weighted most strongly in the example provided herein, other applications may instead weight false positives or false classifications more strongly, and such application-specific preferences can be coded into the optimization process by using a different set of weights assigned to the types of errors. For example, consider the following four examples of different weights:

Table 5: Example weighting schemes for different applications

[0169] By optimizing the score on the training set for the four different weighting schemes shown in Table 5, it was found that the balance of mistakes was shifted away from mistakes being punished most heavily. This effect was demonstrated by examining a breakdown of mistake types across the training set for the four differently tuned detector systems, as shown in Table 6.

Table 6: Distribution of mistake types for detectors tuned using the different weighting schemes shown in Table 5

[0170] The graphical depiction scheme used to generate FIG. 15C may also help illustrate how each of the four detector schemes performs classification of an example image. FIGS. 16A- D represent graphically the results of optimizing automated detector parameters using the four differently weighted score schemes shown in Table 5. It was found that weighting strategies focused on preventing false positives (FIGS. 16A-B) do so by allowing a larger number of false negatives. Conversely, a weighting strategy focused more heavily on preventing false negatives (FIGS. 16C-D) can do so, but (at least in this example) at the expense of allowing two additional false positives errors.

[0171] As should be appreciated from the foregoing, some embodiments are directed to techniques for optimizing parameters of an automated detector. The techniques use a score that characterizes detector performance compared to manual classification of an image and weights that enable the score to punish certain types of mistakes more severely than others. Differential evolution is a numerical optimization technique used in some embodiments to minimize the score on a training set of manually classified images to determine the optimized parameters. Performing differential evolution using differently weighted scores shifts the distribution of mistakes in an expected way for a particular application. FIG. 17 illustrates a process 1700 for tuning parameters of an automated detector in accordance with some embodiments. In act 1710, a first image (or multiple first images) is processed using an automated detector to detect a first set of objects (e.g., cells and other objects) in the first image. Process 1700 then proceeds to act 1712, where detector performance data of the automated detector is determined. As described above, the detector performance data may be based on the first set of objects detected to by the automated detector and a second set of objects detected using manual classification.

Furthermore, the detector performance data may include information about different types of errors produced by the automated detector. Process 1700 then proceeds to act 1714, where the detector parameters of the automated detector are tuned based on the detector performance data. For instance, user input specifying weights for different types of errors may be used in combination with the detector performance data to tune the detector parameters for a particular application. Process 1700 then proceeds to act 1716, where the automated detector having tuned detector parameters is used to process a second image to detect a third set of objects in the second image. Process 1700 may then proceed to act 1718, where information about the detected objects in the third set of objects may be output. For instance, a quantity of objects in the third set may be determined and output.

[0172] As described above, the output produced by a microfluidic system (e.g., the microfluidic system shown in FIG. 2 or 3) after capture, staining, and imaging of microorganisms in a fluid sample may include a grid of multiple (e.g., hundreds or thousands) of image tiles (e.g., fluorescence microscope images) covering the capture area of a microfluidic device within the microfluidic system as an automated x-y stage of the microfluidic system moves the sample to acquire the grid of image tiles. To avoid potentially missing imaging particles (false negatives) in some part of the electrode array due to stage error, the captured image tiles overlap at the borders as shown in FIG. 18. As described in more detail below, defining a set of coordinate transformations allows for matching objects from neighboring images. As shown in FIG. 18, new image tiles are acquired from left to right and top to bottom until the desired area has been imaged. It should be appreciated, however, that the image tiles may be acquired in any order, and embodiments of the present technology are not limited in this respect. As the scan progresses, each newly added tile (center, C) may be compared to its already -processed neighbors for objects in the overlap regions that may have already been detected. Because of the direction the scan progresses, image tiles as neighbors to the northwest, north, northeast, and west (NW, N, NE, W) have already been processed when tile C is obtained. [0173] Due to the overlapping, the same particle of interest (e.g., cell or other microorganism) may appear in two images (for the horizontal and vertical overlap regions) or even four images (for the comer overlap regions). To ensure that an accurate count of the captured particles is performed, some embodiments are directed to a computational solution that accounts for the particles in the overlap regions.

[0174] In some embodiments, the unique contribution of each image tile is identified as the image tile is newly added to the grid of image tiles. Because the image tiles in the grid overlap, all already-processed neighbors of the newly added tile may be checked for objects that have already been detected. In some embodiments, a technique for determining matching coordinates

(e.g., the best matching technique described above applied to automatically detected objects) is used within the overlapping regions to identify objects that have already been detected in a previously processed image tile. The inventors have recognized and appreciated that the coordinates of an object in one image tile may not be directly comparable to the coordinates or same object in a neighboring tile. For example, as shown in FIG. 18, an object in the overlap region between image tiles C and W will be near the left edge of image C and near the right edge of image W. In some embodiments, by measuring the extent of the horizontal and vertical overlaps (oh and o_v) as well as any potential skew due to stage motor error (sh and s_v), a set of coordinate transformations can be defined. These transformations specify what values must be added or subtracted to an object’s coordinates in a neighboring tile (NW, N, NE, W) to directly compare those coordinates to object coordinates in the center image (C). Although the coordinate transformations may not produce exact coordinate matches, the best match technique described herein is configured to identify matches despite small differences in coordinate locations. By identifying particles that have been detected in an overlap region of previously processed image tile, it can be better assured that each particle is only counted once.

[0175] A particular challenge in performing overlap compensation is that particles in the overlap regions may not be detected in all of the overlapping images. For instance, imperfections in the optical system may lead to objects appearing differently depending on, for example, whether they appear in the left or right half of an image. Therefore, a particle in the overlap region may have a somewhat different appearance in each of the overlapping images and may not be detected by the automated detector in all of the images. In some embodiments, using a larger overlap region between images reduces the probability of false negatives due to such a detector failure.

[0176] Accordingly, in some embodiments, the objects detected in a newly added image tile may be checked against the set of all objects that were detected one or more times in the already- processed neighboring image tiles. Drawing inspiration from the mathematics of set theory, this is precisely the set union of the already processed neighboring image tiles. In this framework, the best matching technique can be applied to intersections of images. Therefore, in some embodiments, a union within tolerance between coordinate sets A and B is determined as follows:

• Compute the best matching between object coordinate pairs within an overlap region using the best matching technique described above

• For the calculated object pairings, compute the midpoint between object positions and use it to define the location of shared objects between the sets (that is, the intersection, APB)

• Append to this list the unmatched obj ects from both A and B

[0177] The result will be a set of objects detected in A or B or both, but importantly, the objects detected in both images will only be included in this set (the union) once. Such an approach allows for construction of the exact set of objects against which the newly added tile needs to be compared, the union of already-processed neighbors (U). By computing best matching between object coordinates in the newly added tile (C) and object coordinates in the union of its already-processed neighbors (U), the objects in C that have already been detected in an overlap region can be identified. However, such objects are only counted once, even if they are shared with more than one neighboring image, which is important to obtain accurate counts. Then, the unmatched objects in C contain the unique contribution of that image tile’s contribution to the overall grid of image tiles. By summing up the unique contributions of each image tile, an accurate total count over the entire grid can be determined.

[0178] To evaluate this approach, hand-alignment was performed on a 3x3 grid of images with detected objects annotated by the automated dual-detector system described above, as shown in FIG. 19. The unique objects in this 3x3 grid were then counted, taking care in the overlap regions to only count each object once. This manual count detected 113 cells and 13 other objects. The automated overlap compensation process described above was then performed, which found 114 unique cells and 13 unique other objects, in excellent agreement with the manual count. Simply adding up the counts from each individual tile, 207 cells and 19 other objects were detected, demonstrating the extent to which overcounting can occur due to objects found in the overlap regions, and therefore demonstrating the need for an overlap compensation technique such as the technique described herein.

[0179] As should be appreciated from the foregoing, some microfluidic systems acquire overlapping images to avoid false negatives while imaging a large area of a microfluidic device of the microfluidic system. Cells or other objects in image overlap regions may be detected in any number of the overlapping images. Coordinate transformations based on the optical system stage movement allow for comparison between object locations/coordinates between overlapping images. The best matching technique described herein for comparing object coordinates in an image may be extended to identify matching objects in overlap regions, and the unique contribution of each image tile can be obtained based on the extended best matching technique as described above.

[0180] In the embodiments described above, one or more images captured by an optical system of a microfluidic system are analyzed using a detector configured to apply particular intensity-based segmentation and/or morphological criteria (e.g., size and/or shape criteria) to the image(s) to detect, classify, and/or quantify particles of interest (e.g., cells or other microorganisms) in the image(s). In some embodiments, one or more neural networks are used alone or in combination with an automated detector to detect, classify, and or quantify particles of interest in one or more images. Although the examples of training and using trained neural networks below are described with respect to detecting bacteria, it should be appreciated that the techniques described herein may also be used to detect other microorganisms or particles of interest.

[0181] Some aspects of the technology provide for training neural networks on a panel of selected bacterial species, on indicator bacteria, and/or a broad range of bacterial morphologies to be able to differentiate between debris and microorganisms, where microorganisms may be bacteria, spores, viruses, fungus, mold, yeast, etc. Some aspects of the technology provide for applying the techniques described herein in food safety testing to detect, for example, indicator microorganisms, e.g., E. cob, Salmonella, Listeria etc. Some aspects of the technology provide for applying the techniques described herein for universal capture of microorganisms and artificial intelligence with neural networks trained to recognize specific morphology of indicator bacteria for detection and/or quantification. In some embodiments, the frequency of microorganism capture may be modulated for amplitude and/or frequency. In some embodiments, the techniques described herein may be applied to enhance specificity.

[0182] Aspergillus spores have a very distinctive morphology. Some aspects of the technology provide for use of the techniques described herein to detect and quantify Aspergillus spores (e.g., based on morphology to differentiate them from debris and other microorganisms). In some embodiments, there is provided a database of training data used for specific recognition of a number of microorganisms. For example, the techniques described herein may be applied to detect and quantify a panel of (e.g., 13 or more, or less) indicator water-borne pathogens that cause infections in humans.

[0183] FIG. 20 illustrates a process 2000 for quantifying microorganisms in a sample based on one or more images provided as input to trained neural network, in accordance with some embodiments. In act 2010, an image associated with the sample is provided as input to a trained statistical model (e.g., a trained neural network, a trained machine learning model). The trained statistical model may have been trained to identify microorganisms in an image based on a plurality of labeled images associated with samples including microorganisms and other components. For instance, the images on which the statistical model was trained may include images of one or more electrodes of a microfluidic device (e.g., the microfluidic devices described in connection with the microfluidic systems shown in FIG. 2 or 3) upon which microorganisms are attracted to the surface of the electrode(s), as described herein. Process 2000 then proceeds to act 2020, where the microorganisms in the sample are quantified based, at least in part, on an output (e.g., a labeled image) of the trained statistical model. For instance, the trained statistical model may be trained to one or more microorganisms of interest to be detected, and a number of detected microorganisms may be counted to determine the amount of microorganisms in the sample. Process 2000 then proceeds to act 2020, where the determined quantity of microorganisms in the same is output. For instance, for sterility testing, if even one microorganism (or one live microorganism) is detected in the sample, an indication (e.g., an alarm, a visual indication, or any other suitable indication) that the sample is contaminated may be output.

Example of using trained statistical model to detect microorganisms in sample [0184] For the bacterial cultures E. coli GFP S06 were grown on Nutrient Agar (Teknova) overnight at 37° C. Colonies were picked, resuspended in O.OOlx PBS, and diluted to OD 0.07. The E. coli sample was stained with 1:1000 dilution of SYBR Green I Nucleic Acid Stain (Invitrogen). B. subtilis 66333 (ATCC) were grown on Nutrient Agar (Teknova) overnight at 30° C. Colonies were picked, resuspended in O.OOlx PBS, and diluted to OD 0.06. The B. subtilis sample was stained with 1:1000 dilution of SYBR Green I Nucleic Acid Stain (Invitrogen). For live/dead labelling experiments, B. subtilis were instead incubated with 10 uM SYTO 9 dye and 60 uM propidium iodide from the LIVE/DEAD Bac Light Bacterial Viability Kit (molecular probes).

[0185] For the spore culture, B. subtilis 6633 (ATCC) cells were grown overnight with shaking (200 rpm) at 30° C in nutrient broth (Teknova). Cells were back diluted to OD 0.2 and grown to OD 0.8 before resuspension in Difco sporulation medium (DSM), (8 mg/ml nutrient broth, 1 mg/ml KC1, 1 mM MgS04, 10 uM MnCl₂, 500 uM CaCl₂, 1 mM FeS04). Cells were allowed to sporulate and incubate in DSM for 48 hrs at 30° C with 200 rpm shaking. Before imaging, spores were washed 3x and resuspended in O.OOlx PBS. The unpurified spore sample was stained with 1 : 1000 dilution of SYBR Green I Nucleic Acid Stain (Invitrogen) to label and differentiate remaining vegetative cells from unstained spores.

[0186] Two microliters of the E. coli sample was loaded into a microfluidic channel of a microfluidic device (e.g., the microfluidic device described in connection with the microfluidic system in FIG. 3) and the inlet/outlet holes were sealed with tape to prevent capillary motion.

The device was placed on a SIB V3.0M 2018 BX53 GND No.l stage board and electrical connections were applied using a BRIDGE CP 1.0 14x28_F bridge. The stage board was attached to a SDG 5162 (FGEN-Siglent) function generator and custom LZY-22÷ (LZY-FS-02) amplifier using SMA 24" cables. A 1 MHz, 20 Vpp signal was applied to the electrode to capture bacteria on electrode surface, as described herein. A frequency scan was performed by controlling a function generator to sequentially apply frequencies from 100 MHz to 1 MHz in 1 MHz steps at 4 Vpp.

[0187] The microfluidic stage board was attached to an IX3-SSU automated microscope stage. E. coli were imaged with an optical system that included an Olympus BX63 widefield microscope using a pE-300 ULTRA LED light source (CoolLED) and a 40x objective. SYBR Green stained E. coli and B. subtilis were visualized with 485nm excitation and 510nm emission wavelengths. Additionally, electrode position was imaged with transmitted light bright field microscopy. LIVE/DEAD stained E. coli was visualized with 536 nm excitation and 617 nm emission wavelengths for PI and 485 nm excitation and 510 nm emission wavelengths for SYTO 9. Cells were manually identified (e.g., using manual outlining of the cells) and labeled in a plurality of images acquired by the optical system, and the labeled images were used to train a statistical model. The training data set of images for E. coli detection was generated by loading the sample containing E. coli onto a microfluidic device (e.g., the microfluidic device described in connecting with the microfluidic system of FIG. 3), and acquiring images of the E. coli captured by the electrode(s) of the microfluidic device under the electrical conditions described above. The E. coli bacteria were stained using SYBR Green dye and fluorescent imaging was used to generate fluorescence images, which were annotated as described above. An example image is demonstrated in FIGS. 21A-B. FIG. 21B shows a zoomed in region from the displayed box on the full spiral image displayed in FIG. 21A. The image was acquired with a 40x objective from an E. coli sample captured when a 1MHz, 20 Vpp voltage was applied to the electrode.

[0188] The stability of the algorithm used to implement the trained statistical model was evaluated by providing the same image to the trained statistical model three times and comparing the resulting cell counts each time. This analysis was performed on six images and each run gave the same numerical output for each region of each image, indicating that the algorithm is stable and has a high level of precision, as shown in FIG. 22. FIG. 22 shows bacterial counts for six images run through detection with a trained statistical model three times.

As shown, each time, the resulting cell count output from the model was the same for each region of the image, indicating that the model is stable and reproducible.

[0189] Manual counting was used to assess the accuracy of E. coli cell counts generated by the trained statistical model. Four people independently counted cells on an image, and the manually determined cell counts were compared to cell counts generated by the trained statistical model (labeled as “software”), as shown in FIG. 23. This process was repeated for three different images. FIG. 23 shows manual vs. automated cell counts in accordance with some embodiments. Due to the inherent subjectivity associated with manual counts, there is variability both within the manual counts performed by different people and between manual counts and the automated counts. When the coefficients of variation (CVs) were pooled from three images with four human counts apiece, the total pooled CV was 5.98%. When the automated counts were added to the pooled CVs, the total pooled CV increased to 6.79%, indicating that automated counting is 0.81% more variable than human counts.

[0190] Additionally, automated counting mistakes were evaluated manually and categorized in one of three ways as shown in FIGS. 24 and 25.

1) Merged, when two or more bacteria are counted as one

2) Missed, when a bacterium is not counted by the software

3) Wrong, when an object other than a bacteria is counted as a bacteria

[0191] FIG. 24 shows classifications for different E. coli automated detection errors. Automatically labeled images were checked manually to determine when and how E. coli detection errors were made, with three types of errors above being represented in FIG. 24. FIG. 25 shows manual checks of E. coli automated detection errors. The number of each merge, missing, and wrong errors was compared to the total number of bacteria detected in 25 images.

It was observed that “missing” errors are infrequent, but every image showed both “merge” and “wrong” errors.

[0192] In some embodiments, the trained statistical model to perform automated counting model is configured to process many images from scan experiments, where a sequence of frequencies is applied consecutively, and the bacterial response to each frequency is identified. As shown in FIG. 26, when the bacterial capture results from many images were plotted in a row, a sigmoidal curve was generated, which can be used to identify a crossover frequency (~50 MHz in the example of FIG. 26) from negative dielectrophoresis to positive dielectrophoresis. FIG. 26 show the results of processing 100 image from a frequency scan with the automated counting model. Total bacteria captured by the microfluidic device was determined automatically for each image. [0193] In some embodiments, the trained automated counting model was configured to analyze three channel images, with one bright field and two fluorescence channels. FIGS. 27 A- 27C show the results of the automated counting model update to analyze three channel images in accordance with some embodiments. FIG. 27A shows LIVE/DEAD stained/?, subtilis bright field, Syto-9, and PI channels merged. FIG. 27B shows bacteria detected in the Syto-9 channel. FIG. 27C shows bacteria detected in the PI channel (detected bacteria overlaid with unique color label). Each of the images shown in FIGS. 27A-27C were acquired with a 40x objective from a B. subtilis sample captured at 1 MHz, 20 Vpp with a microfluidic device. In FIG. 27 A, bright field imaging is represented in gray while Syto-9 signal is represented in green and PI signal represented in red. In FIGS. 27B and 27C, detected bacteria are each overlaid with unique color labels and thin white lines represent regions at the edge of the electrode.

[0194] In a further example, the statistical model trained on an E. coli training data set was used to detect bacteria in B. subtilis images to see a model trained to detect one type of microorganism would successfully detect another organism having similar morphology (in this case, another rod-shaped microorganism). Images of SybrGreen stained B. subtilis were processed, and it was observed that the model trained on E. coli images could also successfully identify B. subtilis cells, as shown in FIG. 28A. This same trained model was also applied to FM-464 stained B. subtilis spores and it was observed that the trained model could also successfully identify fluorescently stained/?, subtilis spores as shown in FIG. 28B.

[0195] FIG. 28A shows images displaying B. subtilis vegetative cells detected and labelled by an automated model trained to detect E. coli. The image on right is a zoomed in region from the box on the full spiral image displayed on the left. The image on the left was acquired with a 40x objective from a/?, subtilis sample captured at 1 MHz, 20 Vpp on a microfluidic device. Detected bacteria are each overlaid with unique color labels and thin white lines represent regions at the edge of the electrode. FIG. 28B shows image displaying B. subtilis spores detected and labelled an automated model trained to detect E. coli. The image on right is a zoomed in region from the box on the full spiral image displayed on the left. The image on the left was acquired with a 40x objective from a B. subtilis spore sample captured at 1 MHz, 20 Vpp on a microfluidic device. Detected spores are each overlaid with unique color labels and thin white lines represent the region at the edges of the electrode.

[0196] According to some aspects of the technology described herein, there is provided techniques for optimizing analysis of particles in a fluid sample using different detection techniques (e.g., automated object detector based on intensity and morphological characteristics, trained statistical model (e.g., neural network based) detector). In some embodiments, the output from multiple detectors of the same type (using two different sets of parameters) or of different types (e.g., an intensity/morphology detector and a trained statistical model detector) may be implemented on a single set of data. The results of the two detectors may be combined and compared to obtain a combined result (e.g., a count of microbes in the sample) of increased accuracy.

[0197] As an example, the first set of parameters (for the same or different algorithm) may be chosen in such a way as to recognize all microbes of one type (e.g., bacteria) in the sample and to ensure there are no false negatives. For example, if it is unclear whether the object is a debris particle or a bacterium, then the detector recognizes the object as a bacterium. The result of the detection and subsequent quantification is a number of objects recognized as bacteria Nl. [0198] In this example, the second set of parameters (for the same or different algorithm) may be chosen in such a way as to recognize only bacteria in the sample and to minimize false positives. For example, if it is unclear whether the object is a debris particle or a bacterium, then the detector recognizes the object as debris. The result of the detection and subsequent quantification is a number of objects recognized as bacteria N2.

[0199] The outputs of the detectors (e.g., counts Nl and N2) may be combined in any suitable way. For instance, the positions of objects that were detected differently in both models may be investigated. For example, if there is a bacterium that has two segments the first detector may count it multiple times, whereas the second detector may count it once. In some embodiments, the output of the two detectors may be combined by averaging the two counts Nl and N2 as follows: Naverage = (N1+N2/2) +/- (N2-Nl)/2. In some embodiments, the two counts

Nl and N2 may be associated with different weights such that the average is a weighted average.

In some embodiments, combined output may be configured to ensure zero false negatives.

[0200] The techniques described above for detecting, classifying, and/or quantifying particles of interest (e.g., microorganisms) in a sample take as their starting point one or more images acquired from an optical system of a microfluidic system that includes a microfluidic device configured to capture the particles of interest on one or more electrodes using dielectrophoresis. The inventors have recognized and appreciated that the accuracy of microorganism detection and/or quantification based on images captured by a microfluidic system may be improved by acquiring a negative control scan to be performed prior to processing a sample with the microfluidic device. As described in more detail below, the negative control scan may be used when analyzing the results of sample processing by subtracting the data obtained via the negative control scan from data obtained when processing the sample.

[0201] FIG. 29A schematically illustrates components of a microfluidic system that includes a microfluidic device configured to process fluid samples in accordance with some embodiments. A container 3 having a sample disposed therein may include bacteria 1 (e.g., E. coli. The bacteria may include one species of bacteria or a mix of different species of bacteria. It should be appreciated that although bacteria are described as the microorganisms of interest in this example, other microorganisms may additionally or alternatively be included in the sample and processed with the microfluidic system shown in FIG. 29 A. Container 2 includes a controlled solution (e.g., deionized water, water, buffer solution, sterile buffer, etc.). Container 4 includes a stain. In some embodiments, multiple containers may be used, each of which contains a different stain. Although each of containers 2, 3, and 4 is shown as being coupled to the inlet 5 of a microfluidic passage 6 within a microfluidic device, it should be appreciated that any of containers 2, 3, or 4 may be replaced via a fluid line (e.g., a pipe, a tube, etc.).

[0202] Electrode systems 7, 8 and 9 are electrically coupled to a controller 20 via contact 13, which is configured to provide an electrical signal (e.g., a voltage with amplitude VI and frequency fl) to the electrode systems 7, 8, 9. Controller 20 (e.g., a signal generator) may configured to provide multiple electrical signals (voltages) at multiple amplitudes and frequencies. The electrical signal can be applied to electrodes of opposite polarity as +V1, fl and -VI, fl or as VI, fl and 0V. Multiple signals can be applied to the electrodes. When the electrical signal is applied to the electrode systems 7, 8, 9, those electrode systems are described herein as being “activated.” In response to being activated, electrode systems 7, 8, 9 are configured to generate corresponding electric fields 10, 11, 12. The lines of electric fields 10, 11, 12 also schematically show the trajectories of bacterial motion directed towards the electrodes (without differentiating which area of the electrode) due to the applied electric field and corresponding positive dielectrophoretic force.

[0203] Arrow 14 indicates the direction of fluid flow through the microfluidic passage from the inlet 5 of the microfluidic passage to the outlet 15 of the microfluidic passage. An effluent sample container 16 is coupled to the outlet 15 of the microfluidic passage to collect effluent fluid exiting the microfluidic passage. [0204] The microfluidic system shown in FIG. 29 A also includes an optical system 18 including a fluorescent light source 17 (e.g., an LED) configured to facilitate imaging by optical system 18. For instance, fluorescent light source 17 may be configured to excite a fluorophore attached to a labeled bacteria to capture fluorescent images. In other embodiments, fluorescent light source 17 may be configured to facilitate bright field imaging, for example, when the bacteria in the sample are unlabeled. Optical system 18 may include an optical sensor with a fluorescent light detector, such as a fluorescent microscope or a LED light source with an objective and a detector (e.g., an image sensor). Optical system 18 may be configured to image the electrode system and record the fluorescent signal and/or an image, example of which are described throughout the application. The optical system 18 may be configured to image multiple parts of the electrode system (e.g., to generate a grid of image tiles as discussed in connection with FIG. 18. In some embodiments, the microfluidics device including the microfluidics passage 6 and the electrode systems 7, 8, 9 may be placed on a stage that enables movement in x,y direction to perform a multi-image scan. The microfluidics system shown in FIG. 29A further includes a computer 19 (e.g., at least one hardware processor) configured to process one or more images captured by the optical system 18. For instance, computer 19 may be configured to detect bacteria in fluorescent and/or bright field images and/or quantify bacteria using one or more of the techniques described herein.

[0205] FIG. 30 is a flowchart of a process 3000 for detecting and/or quantifying microorganisms in a sample based, at least in part, on a negative control image in accordance with some embodiments. FIGS. 29B and 29C schematically illustrate some of the acts described in process 3000 and are provided further illustration. With reference to FIG. 29B, in act 3008, a fluid from container 2 (containing a controlled solution) is passed (e.g., pumped) through the microfluidic passage. The electrical signal from the controller 20 may be on or off. When the electrical signal is on, the electric field 10, 11, 12 is generated by the corresponding electrode systems 7, 8, 9. Process 3000 then proceeds to act 3010, where the fluid with stain from container 4 is passed through the microfluidic passage. Process 3000 then proceeds to act 3012, where the fluid from container 2 is again passed through the microfluidic passage to rinse the stain from the microfluidic passage. Process 3000 then proceeds to act 3014, where one or more first “negative scan” images of all or a portion of the surface of electrodes 7, 8, 9 is captured by optical system 18. If there are any microorganisms on the chip causing contamination and/or any other contaminants such as dust, debris or particles 21, they are recorded in the negative scan. [0206] With reference to FIG. 29C, following capture of the negative scan image(s), process 3000 proceeds to act 3016, where the sample in container 3 is passed through the microfluidic passage. As the sample is passed through the microfluidic passage, the electrical signal from the controller 20 is on to generate the electric field 10, 11, 12. In some embodiments, the electrical signal is applied at high amplitude to increase the capture efficiency of bacteria on the electrode systems 7, 8, 9. As shown in FIG. 29C, the bacteria from the sample are shown as moving in the direction 22 towards the electrode system 7 and are being captured on the electrode surface due to a positive dielectrophoretic force. In some embodiments, the entire volume of the sample that needs to be analyzed is passed through the microfluidic passage. If the volume of the sample in the sample container is larger than the volume of the sample that needs to be analyzed, the remaining sample volume may be preserved in the sample container for other purposes. After the sample volume that needs to be analyzed is passed through the microfluidic passage, process 3000 proceeds to act 3018, where the fluid with controlled solution from container 2 is passed through the microfluidic passage to rinse the microfluidic passage. During the rinse, the electric field 10, 11, 12 may remain on to ensure that the microorganisms captured on the electrodes remain captured.

[0207] Process 3000 then proceeds to act 3020, where the fluid with stain from container 4 is passed through the microfluidic passage. Process 3000 then proceeds to act 3022, where the controlled solution is passed through the microfluidic passage to rinse the stain from the electrode systems 7, 8, 9. During the rinse, the electric field remains on to ensure that the microorganisms captured on the electrodes remain captured. In some embodiments that use fluorophore-labeled microorganisms, the amplitude of the electric field may be reduced to avoid affecting the fluorescent signal from stained microorganisms.

[0208] Process 3000 then proceeds to act 3024, where one or more second “positive scan” images of all or a portion of the surface of electrodes 7, 8, 9 is captured by optical system 18. If there are any microorganisms captured by the electrodes, they are recorded in the positive scan. Process 3000 then proceeds to act 3026, where the microorganisms in the sample are detected and/or quantified based on the first negative scan image(s) and the second positive scan image(s). For example, particles that emit fluorescent light (signal) in the negative scan and positive scan image(s) are recognized using one or more of the particle detection techniques described herein. Particles present in the negative scan image(s) are subtracted from particles present in the positive scan image(s). The subtraction may be performed in any suitable way. For instance, in some embodiments, the subtraction is based on particle count. In other embodiments, the subtraction is based on background subtraction of the negative scan images from the positive scan.

[0209] FIG. 31 shows images regarding how negative control scans can reduce the probability of false positives errors in some embodiments. In the images on the left, a negative control scan was performed by processing 4mL of O.OOlxPBS through the microfluidic device under identical flow and electrical conditions used for microbial capture (300pL/min, 1MHz, 50Vpp), followed by lmL of the fluorescent stain SYBR Green I. The electrode area of the microfluidic device was scanned at 20x magnification in bright field and green fluorescence channels. The positive scan images on the right show microscope images after capture of 4mL of E. coli GFP S06 at 100-1000 cells/mL. The capture and imaging was performed using identical conditions to the capture and imaging of the negative control scan, with the only difference being the addition of bacteria to the sample. As shown in the top row of images, some electrodes show no debris in the negative control scan (see image on top left). When such electrodes are analyzed in the positive scan (see image on top right), it can be determined that any microorganisms in the image originate from the analyzed sample. As shown in the bottom row of images, some electrodes do show debris in the negative control scan (see image on bottom left). When such electrodes are analyzed in the positive scan (see image on bottom right), it can be determined that the debris did not originate from the analyzed sample, but instead originated from experimental consumables, environmental contamination or other sources unrelated to the sample being analyzed. In situations where such debris may be mistaken for microorganisms, its presence in the negative control scan may prevent a false positive error. [0210] In some embodiments, two or more stains may be used. A benefit of using two or more stains is only a single integrated scan with two sets of stains (colors) may be captured, instead requiring a separate negative scan and positive scan as described in connection with process 3000. Capturing a single integrated scan may, in some instances, significantly reduce the sample processing time. With reference to FIGS. 29D and 29E, container 4A contains a first stain (stain A) and container 4B contains a second stain (stain B).

[0211] FIG. 32 illustrates a process 3200 for detecting and/or quantifying microorganisms in a sample based on a single image in accordance with some embodiments. FIGS. 29D and 29E schematically illustrate some of the acts described in process 3200 and are provided further illustration. With reference to FIG. 29D, in act 3208, a fluid from container 2 (containing a controlled solution) is passed (e.g., pumped) through the microfluidic passage. The electrical signal from the controller 20 may be on or off. When the electrical signal is on, the electric field 10, 11, 12 is generated by the corresponding electrode systems 7, 8, 9. Process 3200 then proceeds to act 3210, where the fluid with a first stain from container 4 A is passed through the microfluidic passage. The stain A stains particles present in the microfluidic passage in the absence of microorganisms from the actual sample to be processed. Non-limiting examples of such particles include contaminants present in a negative control, e.g., dust, debris, particle, microorganisms. Process 3200 then proceeds to act 3212, where the fluid from container 2 is again passed through the microfluidic passage to rinse the first stain from the microfluidic passage.

[0212] With reference to FIG. 29E, process 3200 then proceeds to act 3214, where the sample in container 3 is passed through the microfluidic passage. As the sample is passed through the microfluidic passage, the electrical signal from the controller 20 is on to generate the electric field 10, 11, 12. In some embodiments, the electrical signal is applied at high amplitude to increase the capture efficiency of bacteria on the electrode systems 7, 8, 9. As shown in FIG. 29E, the bacteria from the sample are shown as moving in the direction 22 towards the electrode systems 7, 8, 9 and are being captured on the electrode surfaces due to a positive dielectrophoretic force. In some embodiments, the entire volume of the sample that needs to be analyzed is passed through the microfluidic passage. If the volume of the sample in the sample container is larger than the volume of the sample that needs to be analyzed, the remaining sample volume may be preserved in the sample container for other purposes. After the sample volume that needs to be analyzed is passed through the microfluidic passage, process 3200 proceeds to act 3016, where the fluid with the second stain in container 4B is passed through the microfluidic passage. The second stain (stain B) is configured to label the microorganisms in the sample. Process 3200 then proceeds to act 3218, where the fluid with the controlled solution from container 2 is passed through the microfluidic passage to rinse the microfluidic passage. During the rinse, the electric field 10, 11, 12 may remain on to ensure that the microorganisms captured on the electrodes remain captured.

[0213] Process 3200 then proceeds to act 3220, where one or more “integrated” images of all or a portion of the surface of electrodes 7, 8, 9 is captured by optical system 18. For example, the optical system may include at least two filters to record the signal from particles stained with stain A and stain B to generate the integrated scan. If there are any microorganisms captured by the electrodes, they are recorded in the integrated scan. Process 3200 then proceeds to act 3222, where the microorganisms in the sample are detected and/or quantified based on the image(s) from the integrated scan. For example, particles that emit fluorescent light (signal) in the integrated scan are recognized using one or more of the particle detection techniques described herein.

[0214] In some embodiments, a third stain, stain C, that is specific to the target microorganisms may be applied. For example, the stain C may be selected such that the stain C does not non-specifically adhere to the surface of debris particles, so that it does not produce a false positive signal from stained non-microorganisms. In some embodiments, a metabolic stain is used to, for example, eliminate the problem of non-specific stain adhesion to debris or nonmicroorganism surfaces as the metabolic stain is configured to enter the inside of cells but not the inside of debris particles. In some embodiments, an autofluorescent signal originating from microorganisms but not from debris particles may be used to detect microorganisms in captured images. Use of autofluorescent signals may lower the false positive rate originating from a fluorescent signal from non-microorganisms that were stained through non-specific surface adhesion.

[0215] In some embodiments, the surface of the microfluidic passage is coated with a surfactant or another substance to limit non-specific surface adhesion of particles other that the microorganisms present in the sample. Such embodiments may lower noise for optical detection with optical system 18. In some embodiments, the techniques described herein are applied to samples that do not have large particles of the size of bacteria, yeast, mold, viruses and or other microorganisms. For instance, such applications include, but are not limited to, microorganism detection and/or quantification from drug substance, from drug product, from water, or from buffer solutions. In some embodiments, bacteria in a sample may be encapsulated in an outer shell to enhance dielectrophoresis capture on the electrode system.

[0216] The techniques described herein may be used to, among other things, rapidly detect and automatically quantify any microorganisms in mammalian cell culture medium from bioreactors (as well as other solutions). Cell culture media from bioreactors typically contain very high number of cells per mL, even up to 10¹⁰/mL. Because of the high concentration of cells per mL, such mediums contain a large number of degraded cells, and thus a large amount of both DNA and RNA nucleic acids released from such cells and free-floating organelles. The presence of such nucleic acids often cause problems for many analytic techniques like PCR, sequencing, etc. The inventors have recognized and appreciated that such nucleic acids (especially in high concentration) also easily stick to the surface of the microfluidic passage of a microfluidic device (e.g., the microfluidic devices described in connection with the microfluidic systems shown in FIG. 2 or 3). Accordingly, if a stain such as Sybr Green I (green fluorescent dye which has the feature of intercalating to nucleic acid molecules) is passed through the microfluidic passage, the entire surface of the microfluidic device will often be stained green during capturing of images by the optical system of the microfluidic system. This effect is undesirable because the nucleic acid molecules covering the entire microfluidic device surface negatively affect: (i) capture of bacteria by coating electrodes on the chip, possibly weakening the electric field; (ii) released captured bacteria from chip for further analysis for e.g. sequencing, bacterial DNA can contain trace amount of mammalian cells nucleic acids which may give a false result; and (iii) automatic detection of bacteria and image analysis of the scanned chip due to the strong green fluorescent background.

[0217] To this end, some embodiments are directed to treatment of mammalian cell culture from bioreactors with commercially available exonuclease (e.g., Pierce Universal Nuclease for Cell Lysis) as a single step reaction before cell culture media are analyzed using a microfluidic device in accordance with the techniques described herein. In a novel approach, exonuclease is used to clarify or remove free-floating nucleic acids by digestion of the nucleic acids directly in mammalian cell culture media. Exonuclease completely digests nucleic acids to oligonucleotides that are less than five bases long, which helps to improve many processes, e.g. enhances filtration of the treated lysate or improves chromatography processing time, etc. Described below are experimental results showing the effectiveness of treating a mammalian cell culture with exonuclease to digest nucleic acids prior to testing the sample with a microfluidic device, in accordance with some embodiments.

[0218] FIG. 33 illustrates a process 3300 for improved quantification of microorganisms in a sample (e.g., from a bioreactor) using a microfluidic device in accordance with some embodiments. In act 3310, the sample is treated with exonuclease to produce a clarified sample.

Process 3300 then proceeds to act 3320, where the clarified sample is passed through the microfluidic passage of a microfluidic device to capture microorganisms in the clarified sample on an electrode surface of the microfluidic device. Process 3300 then proceeds to act 3330, where one or more images of the electrode surface are captured. Process 3300 then proceeds to act 3340, where the captured one or more images are analyzed to quantify an amount of microorganisms in the sample. For instance one or more of the techniques described herein may be used to quantify the amount of microorganism in the sample based on the captured image(s). Example protocols

CHO (Chinese Hamster Ovary) Media [0219] The CHO culture was sampled from a bioreactor, aliquoted to 2- or 15-mL sterile tube and stored in 4°C. Before processing with a microfluidic device and treatment with exonuclease (e.g., Pierce Universal Nuclease), CHO cultured medium was incubated 30 min at room temperature (RT), then diluted 1:100 in a sterile phosphate buffered saline (PBS) pH 7.4 without calcium chloride and magnesium chloride diluted 1:1000 UltraPure Distilled Water to achieve a working range of conductivity. The conductivity of the diluted CHO cultured medium in PBS 1:1000 in DI water was in the range 190-210 µS/cm and was measured at room temperature (RT) using pH/mV/conductivity meter ACCUMET XL200. Aliquots of diluted PBS were stored at 4°C. PBS (FS standard buffer) [0220] All dilutions of CHO processed media were prepared in a sterile phosphate buffered saline (PBS) pH 7.4 w/o calcium chloride and magnesium chloride (Life Technologies, USA) diluted 1:1000 UltraPure Distilled Water (DI water, Life Technologies, USA) in aseptic condition. The conductivity of the PBS 1:1000 in DI water was in the range 19-23 µS/cm and was measured at RT using pH/mV/conductivity meter Accumet® XL200. Aliquots of diluted PBS were stored at 4°C.

Enzymatic reaction

[0221] Pierce Universal Nuclease for Cell Lysis (ThermoFisher Scientific, US A) degrades single-stranded, double-stranded, linear, and circular DNA and RNA and is effective over a wide range of temperatures and pH.

[0222] Nuclease was stored and used according to manufacturer protocol. Briefly, nuclease during sample preparation was transferred from -20°C and kept in ice. A tube was open inside of a biosafety cabinet and 1 μL of nuclease was added directly to 10 mL of RT mammalian cells medium to achieve 25 units/mL final concentration, where one unit corresponds to the amount of enzyme required to produce a change of 1.0 in the absorbance at 260nm of sonicated Herring DNA over 30 minutes at 37°C. To stop enzymatic reaction, the sample was transferred to the ice.

Sample preparation [0223] The CHO culture medium was transferred from 4°C to RT and incubated on the bench for 30 min. Before processing using the microfluidic device, CHO cultured medium was diluted 1:100 in a sterile PBS pH 7.4 diluted 1:1000 UltraPure Distilled Water and was treated with nuclease. Nuclease treatment does not affect conductivity of tested sample as shown in Table 7. All experiments were conducted at room temperature. The prepared sample was ready for processing using a microfluidic device. The control sample was prepared in the same way omitting treatment with nuclease.

Table 7: Conductivity comparison of the nuclease treated and control samples

All work correlated with sample preparation was performed in room temperature and Class II biosafety cabinet.

Microfluidic device

[0224] All tests were conducted on a flow-based microfluidic system (e.g., the microfluidic system shown in FIG. 2) and were performed at RT. To visually evaluate the benefits from nuclease treatment of mammalian cell culture medium from bioreactor and restrain sticking of floating free nucleic acids (DNA, RNA) to the chip surface and electrodes without an electric field, before and after 1 mL of sample, 1 mL of SybrGreen I (1 : 1000) was processed through the flow chip (IMT, Switzerland, for example).

Example 1: Removing of free-floating nucleic acids in CHO cells culture from bioreactor using nuclease

[0225] In this example it was demonstrated that using exonuclease is an efficient method to remove free floating genetic material (DNA, RNA) from bioreactor culture of CHO cells. Two samples were prepared - a first sample without nuclease treatment and a second sample with nuclease treatment. To visualize any background particulates on the microfluidic device surface, 1 mL of SybrGreen I solution in UltraPure DI water was passed through the microfluidic passage of the microfluidic device, with the resultant image shown in FIG. 34A. Then 1 mL of CHO cells culture from bioreactor not treated with nuclease and an additional 1 mL of SybrGreen I solution in UltraPure DI water. Fluorescent visualization based an image captured by the optical system of the microfluidic system revealed a large number of small particulates and large streaks that covered most parts of the imaged microfluidic device as shown in FIG.

34B. After performing a cleaning/sterilizing procedure to confirm that microfluidic device does not contain any biological materials from the process described above, 1 mL of SybrGreen I solution in UltraPure DI water was passed through the microfluidic passage of the microfluidic device and the image in FIG. 34C was captured to visualize background particles. Then, 1 mL of CHO cells culture from bioreactor treated 30 min with nuclease and additional 1 mL of SybrGreen I solution in UltraPure DI water was passed through the microfluidic passage of the microfluidic device. Fluorescent visualization based an image captured by the optical system of the microfluidic system revealed that the large number of small particulates and large streaks that covered most parts of the microfluidic device in the image in FIG. 34B were absent from the image in which the sample was first processed with exonuclease as shown in FIG. 34D. [0226] In this example, it was shown that exonuclease (e.g., Pierce Universal Nuclease for Cell Lysis) treatment of CHO cultured media degraded free-floating genetic material in processed medium, and significantly reduced the green fluorescent background originating from intercalated SybrGreen I to nucleic acids attached to the microfluidic passage surface during sample flow, which may facilitate more precise automatic quantification of microorganisms captured on the electrodes of the microfluidic device.

Example 2. Physico-chemical treatments for removal of surfactants from cells culture medium from bioreactor

[0227] In bioprocesses, due to the introduction of gases into the culture medium, foam occurs and is further stabilized by proteins produced by organisms in the culture. For example, undesired foam formation is observed in bioprocesses used for paper, food, beverage, and drug production. Antifoams can be classified as either hydrophobic solids dispersed in carrier oil, aqueous suspensions/emulsions, liquid single components or solids and may contain surfactants.

Nonionic surfactant (e.g., Polysorbate 80) may be used as a neutralizing agent to neutralize activity of the antimicrobial substances such as quaternary ammonium compounds (QACs), iodine, and parabens according to USP <61> Microbiological examination of nonsterile

Products: microbial enumeration tests. However, the presence of the surfactants in cells culture medium analyzed with a microfluidic device using one or more of the techniques described herein can be problematic due to bubbles or even unstable foam forming when the sample is passed through the microfluidic passage of the microfluidic device. For instance, bubbles or light/sporadic foaming can affect the bacteria capture process/efficiency, resulting in inaccurate results. To this end, some embodiments provide for adsorption of surfactants of mammalian or cell culture from bioreactors onto activated carbon before culture media are analyzed using a microfluidic device.

Examples and General Protocols CHO (Chinese Hamster Ovary) Media [0228] The CHO culture was sampled from bioreactor, aliquoted to 2- or 15-mL sterile tube and stored in 4°C. The conductivity of the cell culture medium was in the range 10-12 mS/cm and was measured at RT using pH/mV/conductivity meter Accumet® XL200. PBS (FS standard buffer) [0229] All dilutions of CHO processed media were prepared in a sterile phosphate buffered saline (PBS) pH 7.4 without calcium chloride and magnesium chloride (Life Technologies, USA) diluted 1:1000 UltraPure Distilled Water (DI water, Life Technologies, USA) in aseptic condition. The conductivity of the PBS 1:1000 in DI water was in the range 19-23 µS/cm and was measured at RT using pH/mV/conductivity meter Accumet® XL200. Aliquots of diluted PBS were stored at 4°C.

Activated charcoals

[0230] 1. Activated carbon DARCO (Sigma, USA) - 100 mesh was used in a powder form.

[0231] 2. Activated carbon was used in a powder form (Sigma, USA).

Sample preparation

[0232] The CHO culture medium was transferred from 4°C to RT and incubated on the bench for 30 min. Before processing using the microfluidic device, CHO cultured medium was incubated with active charcoal for 30 min in RT. Charcoal was separated from cell culture medium by filtration or decanted, then the microorganism was spiked. All experiments were conducted at RT. Such a prepared sample was then ready for processing using the microfluidic device. A control sample was prepared in the same way omitting incubation with activated charcoal.

[0233] All work correlated with sample preparation can be performed in room temperature and Class II biosafety cabinet.

Microfluidic device [0234] All tests were conducted on a flow-based microfluidic system (e.g., the microfluidic system shown in FIG. 2) and were performed at RT. To evaluate the benefits from incubation of mammalian cell culture medium from bioreactor with activated charcoal, filtrated/decanted media with spiked microorganism were processed through the microfluidic device, visualized, and quantified using the microfluidic system shown in FIG. 2.

[0235] In this example it was shown that treatment of a sample with activated charcoal prior to processing the sample using a microfluidic device and one or more of the techniques described herein, reduced the presence of foam in the sample, which may facilitate more precise automatic quantification of microorganisms captured on the electrodes of the microfluidic device.

[0236] According to some aspects of the technology described herein, there is provided a screening method for screening for contamination in sterile manufacturing.

[0237] Use of a microfluidic device according to the techniques described herein enables 100% or near 100% capture of all bacteria at low bacterial concentrations. In addition, the data analysis techniques described herein provide for minimization or elimination of false negative errors in the analysis. The screening techniques described in more detail below provide an approach to screening which reduces the number of tests that are necessary to be performed by a bacterial culture, thereby significantly reducing the amount of time necessary to perform screening.

Example Screening Applications

[0238] Multiple industrial processes (e.g., manufacturing of injectable drugs, biologies and diagnostic tests such as urine tests for urinary tract infection diagnosis, blood tests for sepsis diagnosis, and spinal cord fluid tests) require testing samples for the presence of microorganisms. Testing for the presence of microorganisms, when based on bacterial cultures, takes multiple hours or even days and is labor intensive. Sterility testing may take 14 days and bioburden testing using traditional methods requires growing bacterial colonies for 3-5 days.

The United States Pharmacopeia (USP) regulates sterility testing and bioburden testing in the pharmaceutical industry.

[0239] Testing fluid samples for sterility requires detecting the presence of even one microorganism, according to the USP. The sterility test requires detecting any type of bacterial contamination, yeast, mold and fungus. Conventional sterility tests are complicated and labor intensive. For instance, sterility testing according to typical methods requires specialized laboratory infrastructure, trained personnel and is prone to human error. For in-process control in biomanufacturing, bioburden tests are used with an expectation that no colonies grow on plates. Sterility testing only requires detection and does not require quantification.

[0240] According to some aspects of the technology described herein, there are provided improved screening techniques that enable real-time information about the presence of microorganisms in a sample in a shorter amount of time (e.g., less than 8 hours) than conventional sterility testing technique. This information may be used to determine if there is microbial contamination present in a sample, and may be used, for example, for real time drug release (RTR) testing. This information may alternatively be used for final product release testing, for informing decisions about the next steps in product manufacturing and/or for use in process control.

[0241] FIG. 35 illustrates a process 3500 for performing a sterility test on a sample in accordance with some embodiments. In act 3510, a first portion of the sample is passed through a microfluidic passage of a microfluidic system. Process 3500 then proceeds to act 3520, where a dielectrophoretic force is generated within the microfluidic passage as the sample is passed through the microfluidic passage, wherein the dielectrophoretic force acts on the first portion of the sample. Process 3500 then proceeds to act 3530, where based on the response of the first portion of the sample to the dielectrophoretic force, it is determined whether the first portion of the sample includes microorganisms. Process 3500 then proceeds to act 3540, where the sample is labeled as sterile when it is determined in act 3530 that the first portion of the sample does not include microorganisms. In some embodiments, the microfluidic system that includes the microfluidic device is designed in such a way that the false negative rate is zero, for example, at least in part due to the 100% capture efficiency of the microfluidic system and/or use of the data analysis techniques described herein.

[0242] Some aspects provide for screening for contamination in manufacturing processes based on microorganisms capture on electrodes and optical inspection. Some aspects provide for screening for disease agents in human samples based on microorganisms capture on electrodes and optical inspection. Based on the use of a microfluidic system to perform screening, this approach allows for eliminating a significant amount of bacterial culture plates and thereby reduces the testing time and cost, for example, in an amount equivalent to the number of samples that are able to be processed without bacterial culture. The microfluidic system may be operated and/or data obtained from the system may be analyzed, in a manner that eliminates false negative results (e.g., the failure to detect the presence of a bacteria). This optimization may be at the cost of increasing the number of false positive results. Given this optimization, if the result of the test is negative, the sample is considered to contain zero microorganisms and does not require further sterility or bioburden testing. A negative test result may then be a basis for real time drug release.

[0243] As the microfluidic system and/or data analysis applied are optimized to eliminate false negative results, some number of false positive results may be obtained. A false positive result may be subsequently tested in the customary manner (e.g., with a bacterial culture). Because a significant number of samples will return no false positives, the method enables the elimination of the need for performing bacterial cultures of a large number of samples. Accordingly, the amount of time needed to perform screening is greatly reduced. For example, where the false positive rate is not higher than 50%, a reduction by a factor of two of the time and testing costs for the manufacturer can be achieved.

[0244] As shown in FIG. 36, some embodiments related to sterility testing may include the following outcomes: 1) The apparatus processes the sample. If the test result is negative, that means the sample is free from contamination. The sample may be considered ready for real time drug release. 2) If the test result is positive, that sample is either contaminated or the test result is a false positive. To determine which is the case, the remainder of the unprocessed original sample is processed using a conventional technique or another method.

[0245] In some embodiments, an optical system may be configured to use different magnification objectives for detecting, quantifying and/or classifying microorganisms using a microfluidic system that includes a microfluidic device, examples of which are described herein. For example, to achieve a faster turnaround time to analyze images, a lower magnification objective, which allows for faster scanning of the chip area but is sufficient to detect microorganisms and quantify them may be used. After the microorganisms have been detected, a higher magnification objective of the optical system may be used to selectively scan spots that were detected using the lower magnification objective. In this way both high magnification and high resolution images are collected, which allows for, e.g., differentiation of microorganisms from debris and non-microorganism particles, and/or identification of bacteria and microorganisms based on their morphology, e.g., for identifying contamination in biopharma manufacturing. [0246] Using this approach, a database of the most common microorganisms that cause contamination may be generated and used for further processing of samples as follows:

1) A sample (e.g., associated with a drug) is tested with a microfluidic system using one or more of the techniques described herein.

2) When no contamination (e.g., microorganisms) is detected in the sample, the sample is considered to be clean and the drug can be released.

3) When contamination is detected in the sample, the captured microorganisms can be scanned with higher magnification and the resulting images can be identified to see if they match any of the images in the database in an attempt to identify the microorganisms causing the contamination.

4) If a match to the images in the database is not found, then other methods (e.g., plating) may be used to identify the microorganisms causing contamination.

[0247] In some embodiments, the techniques described herein may be used for improved bioburden testing. FIG. 37 illustrates a process 3700 for determining whether a sample is in condition for release (e.g., for pharmaceutical manufacturing). In act 3710, a first portion of the sample is passed through a microfluidic passage of a microfluidic system. Process 3700 then proceeds to act 3720, where a dielectrophoretic force is generated within the microfluidic passage as the sample is passed through the microfluidic passage, wherein the dielectrophoretic force acts on the first portion of the sample. Process 3700 then proceeds to act 3730, where a concentration of microorganisms in the first portion of the sample as immobilized on the surface of the electrode due to the dielectrophoretic force is determined. Process 3700 then proceeds to act 3740, where it is determined whether the concentration of microorganisms is less than a threshold concentration. If it is determined in act 3740 that the concentration of microorganisms is not less than the threshold value, process 3700 proceeds to act 3750 where it is determined that the sample is not in condition for release. Otherwise, if it is determined in act 3740 that the concentration of microorganisms is less than the threshold concentration, process 3700 proceeds to act 3760, where it is determined that the sample is in condition for release.

[0248] Having thus described several aspects and embodiments of the technology set forth in the disclosure, it is to be appreciated that various alterations, modifications, and improvements will readily occur to those skilled in the art. [0249] Such alterations, modifications, and improvements are intended to be within the spirit and scope of the technology described herein. For example, those of ordinary skill in the art will readily envision a variety of other means and/or structures for performing the function and/or obtaining the results and/or one or more of the advantages described herein, and each of such variations and/or modifications is deemed to be within the scope of the embodiments described herein. Those skilled in the art will recognize or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments described herein. It is, therefore, to be understood that the foregoing embodiments are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, inventive embodiments may be practiced otherwise than as specifically described. In addition, any combination of two or more features, systems, articles, materials, kits, and/or methods described herein, if such features, systems, articles, materials, kits, and/or methods are not mutually inconsistent, is included within the scope of the present disclosure.

[0250] The above-described embodiments can be implemented in any of numerous ways.

One or more aspects and embodiments of the present disclosure involving the performance of processes or methods may utilize program instructions executable by a device (e.g., a computer, a processor, or other device) to perform, or control performance of, the processes or methods. In this respect, various inventive concepts may be embodied as a computer readable storage medium (or multiple computer readable storage media) (e.g., a computer memory, one or more floppy discs, compact discs, optical discs, magnetic tapes, flash memories, circuit configurations in Field Programmable Gate Arrays or other semiconductor devices, or other tangible computer storage medium) encoded with one or more programs that, when executed on one or more computers or other processors, perform methods that implement one or more of the various embodiments described above. The computer readable medium or media can be transportable, such that the program or programs stored thereon can be loaded onto one or more different computers or other processors to implement various ones of the aspects described above. In some embodiments, computer readable media may be non-transitory media.

[0251] The terms “program” or “software” are used herein in a generic sense to refer to any type of computer code or set of computer-executable instructions that can be employed to program a computer or other processor to implement various aspects as described above.

Additionally, it should be appreciated that according to one aspect, one or more computer programs that when executed perform methods of the present disclosure need not reside on a single computer or processor but may be distributed in a modular fashion among a number of different computers or processors to implement various aspects of the present disclosure.

[0252] Computer-executable instructions may be in many forms, such as program modules, executed by one or more computers or other devices. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically, the functionality of the program modules may be combined or distributed as desired in various embodiments.

[0253] Also, data structures may be stored in computer-readable media in any suitable form. For simplicity of illustration, data structures may be shown to have fields that are related through location in the data structure. Such relationships may likewise be achieved by assigning storage for the fields with locations in a computer-readable medium that convey relationship between the fields. However, any suitable mechanism may be used to establish a relationship between information in fields of a data structure, including through the use of pointers, tags or other mechanisms that establish relationship between data elements.

[0254] The above-described embodiments of the present technology can be implemented in any of numerous ways. For example, the embodiments may be implemented using hardware, software or a combination thereof. When implemented in software, the software code can be executed on any suitable processor or collection of processors, whether provided in a single computer or distributed among multiple computers. It should be appreciated that any component or collection of components that perform the functions described above can be generically considered as a controller that controls the above-described function. A controller can be implemented in numerous ways, such as with dedicated hardware, or with general purpose hardware (e.g., one or more processor) that is programmed using microcode or software to perform the functions recited above and may be implemented in a combination of ways when the controller corresponds to multiple components of a system.

[0255] Further, it should be appreciated that a computer may be embodied in any of a number of forms, such as a rack-mounted computer, a desktop computer, a laptop computer, or a tablet computer, as non-limiting examples. Additionally, a computer may be embedded in a device not generally regarded as a computer but with suitable processing capabilities, including a Personal Digital Assistant (PDA), a smartphone or any other suitable portable or fixed electronic device. [0256] Also, a computer may have one or more input and output devices. These devices can be used, among other things, to present a user interface. Examples of output devices that can be used to provide a user interface include printers or display screens for visual presentation of output and speakers or other sound generating devices for audible presentation of output. Examples of input devices that can be used for a user interface include keyboards, and pointing devices, such as mice, touch pads, and digitizing tablets. As another example, a computer may receive input information through speech recognition or in other audible formats.

[0257] Such computers may be interconnected by one or more networks in any suitable form, including a local area network or a wide area network, such as an enterprise network, and intelligent network (IN) or the Internet. Such networks may be based on any suitable technology and may operate according to any suitable protocol and may include wireless networks, wired networks or fiber optic networks.

[0258] Also, as described, some aspects may be embodied as one or more methods. The acts performed as part of the method may be ordered in any suitable way. Accordingly, embodiments may be constructed in which acts are performed in an order different than illustrated, which may include performing some acts simultaneously, even though shown as sequential acts in illustrative embodiments.

[0259] All definitions, as defined and used herein, should be understood to control over dictionary definitions, definitions in documents incorporated by reference, and/or ordinary meanings of the defined terms.

[0260] The indefinite articles “a” and “an,” as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean “at least one.” [0261] The phrase “and/or,” as used herein in the specification and in the claims, should be understood to mean “either or both” of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with “and/or” should be construed in the same fashion, i.e., “one or more” of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the “and/or” clause, whether related or unrelated to those elements specifically identified. Thus, as anon-limiting example, a reference to “A and/or B”, when used in conjunction with open-ended language such as “comprising” can refer, in one embodiment, to A only (optionally including elements other than B); in another embodiment, to B only (optionally including elements other than A); in yet another embodiment, to both A and B (optionally including other elements); etc.

[0262] As used herein in the specification and in the claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, “at least one of A and B” (or, equivalently, “at least one of A or B,” or, equivalently “at least one of A and/or B”) can refer, in one embodiment, to at least one, optionally including more than one, A, with no B present (and optionally including elements other than B); in another embodiment, to at least one, optionally including more than one, B, with no A present (and optionally including elements other than A); in yet another embodiment, to at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and optionally including other elements); etc.

[0263] Also, the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. The use of "including," "comprising," or "having," “containing,” “involving,” and variations thereof herein, is meant to encompass the items listed thereafter and equivalents thereof as well as additional items.

[0264] In the claims, as well as in the specification above, all transitional phrases such as “comprising,” “including,” “carrying,” “having,” “containing,” “involving,” “holding,” “composed of,” and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases “consisting of’ and “consisting essentially of’ shall be closed or semi-closed transitional phrases, respectively.

[0265] The terms “substantially”, “approximately”, and “about” may be used to mean within

±20% of a target value in some embodiments, within ±10% of a target value in some embodiments, within ±5% of a target value in some embodiments, within ±2% of a target value in some embodiments. The terms “approximately” and “about” may include the target value.

[0266] Use of ordinal terms such as “first,” “second,” “third,” etc., in the claims to modify a claim element does not by itself connote any priority, precedence, or order of one claim element over another or the temporal order in which acts of a method are performed, but are used merely as labels to distinguish one claim element having a certain name from another element having a same name (but for use of the ordinal term) to distinguish the claim elements.

Claims

1. A method for classifying a particle in an image as belonging to a first group of particles, the particle being represented in the image as a plurality of contiguous pixels of the image, the method comprising: determining whether an intensity value for the particle in the image is greater than a first threshold value; determining whether at least one morphological characteristic of the particle in the image satisfies a first set of criteria associated with particles in the first group of particles; classifying the particle as belonging to the first group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image satisfies the first set of criteria; and outputting a result of the classifying of the particle as belonging to the first group of particles.

2. The method of claim 1, wherein the morphological characteristic comprises a size and/or shape of the particle.

3. The method of claim 2, wherein the size of the particle comprises a number of contiguous pixels representing the particle in the image.

4. The method of claim 2, wherein the size of the particle is determined based on at least one measurement selected from the group consisting of an area, a circumference, a diameter, an extent, and a bisector.

5. The method of claim 2, wherein the shape of the particle is determined based on at least one measurement selected from the group consisting of an aspect ratio, an elongation value, a convexity value, a shape factor, and a sphericity value.

6. The method of claim 1, wherein the particle comprises one of a microorganism or debris.

7. The method of claim 6, wherein the microorganism comprises one of bacteria, yeast, or mold. 8 The method of claim 1, wherein the first group of particles comprise microorganisms.

9. The method of claim 1, wherein the image includes particles belonging to the first group of particles and particles belonging to a second group of particles, the method further comprising: classifying the particle as belonging to the second group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image does not satisfy the first set of criteria.

10. The method of claim 9, wherein the second group of particles comprises debris.

11. The method of claim 9, wherein the second group of particles comprises components other than microorganisms.

12. The method of claim 1, further comprising: when the particle is classified as belonging to the first group of particles, further classifying the particle as belonging to a first subgroup or a second subgroup of the first group of particles at least in part by: determining whether the value for the at least one morphological characteristic satisfies a second set of criteria; and classifying the particle as belonging to the first subgroup when the at least one morphological characteristic satisfies the second set of criteria.

13. The method of claim 12, wherein the first subgroup of the first group of particles comprises a first type of bacteria and the second subgroup of the first group of particles comprises a second type of bacteria different than the first type of bacteria.

14. The method of claim 1, wherein outputting a result of the classifying of the particle as belonging to the first group of particles comprises: generating an overlay visualization indicating that the particle belongs to the first group of particles; and displaying the overlay visualization and the image on a display.

15. The method of claim 1, wherein the image is a fluorescence image.

16. The method of claim 1, further comprising: quantifying a number of particles classified as belonging to the first group of particles; and outputting the number of particles classified as belonging to the first group of particles.

17. The method of claim 1, further comprising: inputting a sample comprising the particle into a microfluidic passage of a microfluidic device; immobilizing the particle onto a surface of an electrode disposed in the microfluidic passage with at least one dielectrophoretic force; and capturing the image while the particle is immobilized on the surface of the electrode.

18. A system for classifying a particle in an image as belonging to a first group of particles, the particle being represented in the image as a plurality of contiguous pixels of the image, the system comprising: a microfluidic passage for receiving a sample, the sample comprising the particle; at least one electrode disposed in the microfluidic passage, the at least one electrode configured to immobilize, when activated, the particle onto a surface of the at least one electrode using dielectrophoresis; an optical system configured to capture the image while the particle is immobilized on the surface of the at least one electrode; and at least one computing device configured to: determine, based on the image obtained by the optical device, whether an intensity value for the particle in the image is greater than a first threshold value; determine whether at least one morphological characteristic of the particle in the image satisfies a set of criteria associated with particles in the first group of particles; classify the particle as belonging to the first group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image satisfies the set of criteria; and output a result of the classifying of the particle as belonging to the first group of particles.

19. The system of claim 18, wherein the morphological characteristic comprises a size and/or shape of the particle.

20. The system of claim 19, wherein the size of the particle comprises a number of contiguous pixels representing the particle in the image.

21. The system of claim 19, wherein the size of the particle is determined based on at least one measurement selected from the group consisting of an area, a circumference, a diameter, an extent, and a bisector.

22. The system of claim 19, wherein the shape of the particle is determined based on at least one measurement selected from the group consisting of an aspect ratio, an elongation value, a convexity value, a shape factor, and a sphericity value.

23. The system of claim 18, wherein the particle comprises one of a microorganism or debris.

24. The system of claim 23, wherein the microorganism comprises one of bacteria, yeast, or mold.

25. The system of claim 18, wherein the first group of particles comprise microorganisms.

26. The system of claim 18, wherein the image includes particles belonging to the first group of particles and particles belonging to a second group of particles, the at least one computing device being further configured to: classify the particle as belonging to the second group of particles when the intensity value for the particle in the image is greater than the threshold value and the at least one morphological characteristic of the particle in the image does not satisfy the first set of criteria.

27. The system of claim 26, wherein the second group of particles comprises debris.

28. The system of claim 26, wherein the second group of particles comprises components other than microorganisms.

29. The system of claim 18, wherein: when the particle is classified as belonging to the first group of particles, the at least one computing device is further configured to classify the particle as belonging to a first subgroup or a second subgroup of the first group of particles at least in part by: determining whether the value for the at least one morphological characteristic satisfies a second set of criteria; and classifying the particle as belonging to the first subgroup when the at least one morphological characteristic satisfies the second set of criteria.

30. The system of claim 29, wherein the first subgroup of the first group of particles comprises a first type of bacteria and the second subgroup of the first group of particles comprises a second type of bacteria different than the first type of bacteria.

31. The system of claim 18, wherein outputting a result of the classifying of the particle as belonging to the first group of particles comprises: generating an overlay visualization indicating that the particle belongs to the first group of particles; and displaying the overlay visualization and the image on a display.

32. The system of claim 18, wherein the image is a fluorescence image.

33. The system of claim 18, wherein the at least one computing device is further configured to: quantify a number of particles classified as belonging to the first group of particles; and output the number of particles classified as belonging to the first group of particles.

34. A method for tuning detector parameters of an automated detector used to automatically detect microorganisms in images, the method comprising: processing a plurality of first images with the automated detector to detect a first set of objects in each of the plurality of first images, each first set of objects including particles of interest and other objects; determining detector performance data of the automated detector based on the first set of objects for each of the plurality of first images and a corresponding second set of objects determined from manual identification of objects in each of the plurality of first images, the detector performance data including information about different types of errors generated by the automated detector; tuning the detector parameters of the automated detector based, at least in part, on the detector performance data; processing a second image with the automated detector having tuned detector parameters to detect a third set of objects in the second image; and outputting information about the detected objects in the second image.

35. The method of claim 34, wherein the third set of objects includes the particles of interest and other objects, and wherein the method further comprises: quantifying an amount of particles of interest in the third set of objects.

36. The method of claim 34, wherein determining detector performance data of the automated detector based on the first set of objects for each of the plurality of first images and a corresponding second set of objects determined from manual identification of objects in each of the plurality of first images comprises comparing object coordinates of objects in the first set of objects and objects in the second set of objects to identify objects detected in both the first and second sets for each of the plurality of first images.

37. The method of claim 36, wherein an object is identified to be detected in both the first and second sets when a distance between the object coordinates for the object in the first set and the object coordinates for the object in the second set are less than a threshold distance apart.

38. The method of claim 36, wherein determining detector performance data further comprises: computing a pairwise distance matrix between the object coordinates in the first set of objects and object coordinates in the second set of objects; and identifying, based on the computed pairwise distance matrix, objects with a matching set of object coordinates.

39. The method of claim 38, wherein identifying objects with a matching set of coordinates comprises using a Hungarian algorithm to identify the objects with a matching set of coordinates.

40. The method of claim 34, wherein tuning the detector parameters of the automated detector is further based, at least in part, on user input assigning weights assigned to each of the different types of errors generated by the automated detector.

41. A method for determining a quantity of microorganisms in a set of image tiles, the method comprising: receiving a set of image tiles, the set including a first image tile and a second image tile having an overlap region with the first image tile; detecting, with an automated detector, a first set of microorganisms within the first image tile and a second set of microorganisms within the second image tile, wherein each of the microorganisms in the first set is associated with first object coordinates and each of the of microorganisms in the second set is associated with second object coordinates; identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region; determining the quantity of microorganisms in the set of image tiles based on a number of microorganisms in the first set of microorganisms, a number of microorganisms in the second set of microorganisms and a number of object coordinate pairs identified in the overlap region; and outputting the quantify of microorganisms in the set of image tiles.

42. The method of claim 41, wherein identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region comprises identifying an object coordinate pair when the first object coordinates and the second object coordinates are less than a threshold distance apart.

43. The method of claim 41, wherein identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region comprises: transforming the first object coordinates into a common coordinate space with the second object coordinates; and identifying an object coordinate pair based on the transformed first object coordinates and the second object coordinates.

44. The method of claim 41, wherein identifying, based on first object coordinates and the second object coordinates, object coordinate pairs within the overlap region comprises using a Hungarian algorithm to identify the object coordinate pairs.

45. A method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the method comprising: identifying, based on an image of the sample and a first set of criteria including a first intensity threshold and one or more first morphological criteria, a first set of objects in the sample that have an intensity value in the image above the first intensity threshold and that have a shape characteristic and/or a size characteristic in the image that satisfies the one or more first morphological criteria, the objects in the first set of objects being classified as belonging to a first group of particles associated with the microorganisms of the sample; identifying, based on the image of the sample and a second set of criteria including a second threshold intensity and/or one or more second morphological criteria, a second set of objects in the sample that have an intensity value in the image above the intensity threshold and/or that have a shape characteristic and/or a size characteristic in the image that satisfies the one or more second morphological criteria, the objects in the second set of objects being classified as belonging to the first group of particles associated with the microorganisms of the sample, determining, as a first count, a number of objects in the first set of objects that belong to the first group of particles; determining, as a second count, a number of objects in the second set of objects that belong to the first group of particles; determining the quantity of microorganisms in the sample based on a combination of the first count and the second count; and outputting the determined quantity of microorganisms in the sample.

46. The method of claim 45, wherein determining the quantity of microorganisms in the sample based on a combination of the first count and the second count comprises determining an average of the first count and the second count.

47. The method of claim 46, wherein determining an average of the first count and the second count comprises determining a weighted average of the first count and the second count.

48. The method of claim 45, wherein the first set of criteria is selected to minimize false positive detection errors.

49. A method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the method comprising: identifying, based on an image of the sample and a first set of criteria including a first intensity threshold and one or more first morphological criteria, a first set of objects in the sample that have an intensity value in the image above the first intensity threshold and that have a shape characteristic and/or a size characteristic in the image that satisfies the one or more first morphological criteria, the objects in the first set of objects being classified as belonging to a first group of particles associated with the microorganisms of the sample; providing, as input to a trained statistical model, the image of the sample, wherein the trained statistical model was trained to identify microorganisms in an image based on a plurality of labeled images of samples including microorganisms and other components; determining the quantity of microorganisms in the sample by: determining, as a first count, a number of objects in the first set of objects that belong to the first group of particles; determining, as a second count, a number of labeled objects identified in a labeled image output from the trained statistical model; and determining the quantity of microorganisms in the sample based on a combination of the first count and the second count.

50. The method of claim 49, wherein determining the quantity of microorganisms in the sample based on a combination of the first count and the second count comprises determining an average of the first count and the second count.

51. The method of claim 49, wherein determining an average of the first count and the second count comprises determining a weighted average of the first count and the second count.

52. The method of claim 49, wherein the first set of criteria is selected to minimize false positive detection errors.

53. A method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the method comprising: providing, as input to a trained statistical model, an image of the sample, wherein the trained statistical model was trained to identify microorganisms in an image based on a plurality of labeled images of samples including microorganisms and other components; determining based on a labeled image output from the trained statistical model, the quantity of the microorganisms in the sample; and outputting the determined quantity of microorganisms in the sample.

54. The method of claim 53, wherein the trained statistical model was trained to identify a first type of microorganisms in an image, the image provided as input to the trained statistical model is an image of a sample including a second type of microorganisms, and determining the quantity of microorganisms in the sample comprises determining the quantity of the second type of microorganisms in the sample.

55. The method of claim 54, wherein the image is a fluorescent image.

56. The method of claim 53, wherein the image is a multi-channel image.

57. The method of claim 56, wherein the multi-channel image includes a bright field channel and two fluorescent channels.

58. The method of claim 53, wherein the trained statistical model includes at least one trained neural network.

59. The method of claim 53, wherein the microorganisms are Aspergillus spores.

60. The method of claim 53, wherein the microorganisms in the sample include a first type of microorganisms and a second type of microorganisms, the image provided as input to the trained statistical model includes both the first type of microorganisms and the second type of microorganisms, and the trained statistical model is trained to label in the output image, only the first type of microorganisms and not the second type of microorganisms.

61. The method of claim 60, wherein the first type of microorganisms have a different morphology than the first type of microorganisms.

62. A system for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the system comprising: a microfluidic passage for receiving the sample; at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, particles of the sample on a surface of the at least one electrode using dielectrophoresis; an optical system configured to capture at least one image of the surface of the at least one electrode; and at least one computing device configured to: provide, as input to a trained statistical model, an image of the sample, wherein the trained statistical model was trained to identify microorganisms in an image based on a plurality of labeled images of samples including microorganisms and other components; determine based on a labeled image output from the trained statistical model, the quantity of the microorganisms in the sample; and output the determined quantity of microorganisms in the sample.

63. The system of claim 62, wherein the trained statistical model was trained to identify a first type of microorganisms in an image, the image provided as input to the trained statistical model is an image of a sample including a second type of microorganisms, and determining the quantity of microorganisms in the sample comprises determining the quantity of the second type of microorganisms in the sample.

64. The system of claim 62, wherein the image is a fluorescent image.

65. The system of claim 62, wherein the image is a multi-channel image.

66. The system of claim 65, wherein the multi-channel image includes a bright field channel and two fluorescent channels.

67. The system of claim 62, wherein the trained statistical model includes at least one trained neural network.

68. The system of claim 62, wherein the microorganisms are Aspergillus spores.

69. The system of claim 62, wherein the microorganisms in the sample include a first type of microorganisms and a second type of microorganisms, the image provided as input to the trained statistical model includes both the first type of microorganisms and the second type of microorganisms, and the trained statistical model is trained to label in the output image, only the first type of microorganisms and not the second type of microorganisms.

70. The system of claim 69, wherein the first type of microorganisms have a different morphology than the first type of microorganisms.

71. A method for improved detection and/or quantification of microorganisms in a sample, the sample including microorganisms and other components, the method comprising: passing a first fluid comprising a stain through a microfluidic passage of a microfluidic device, the microfluidic passage including at least one electrode disposed therein; passing a second fluid through the microfluidic passage after passing the first fluid through the microfluidic passage, the second fluid comprising a controlled solution without the stain; capturing, with an optical system, a first image including a surface of the at least one electrode after passing the second fluid through the microfluidic passage; passing the sample through the microfluidic passage after capturing the first image, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the sample is passed through the microfluidic passage; passing the first fluid through the microfluidic passage while the microorganisms remain captured on the surface of the at least one electrode; passing the second fluid through the microfluidic passage after passing the first fluid through the microfluidic passage and while the microorganisms remain captured on the surface of the at least one electrode; capturing, with the optical system, a second image of the surface of the at least one electrode; and detecting and/or quantifying microorganisms in the sample based on the first image and the second image.

72. The method of claim 71, further comprising: passing the second fluid through the microfluidic passage prior to passing the first fluid through the microfluidic passage prior to capturing the first image.

73. The method of claim 71, further comprising: passing the second fluid through the microfluidic passage prior to passing the first fluid through the microfluidic passage and after passing the sample through the microfluidic passage.

74. The method of claim 71, wherein passing the first fluid, the second fluid, or the sample through the microfluidic passage comprises pumping the first fluid, the second fluid, or the sample through the microfluidic passage.

75. The method of claim 71, further comprising: applying at least one first voltage to the at least one electrode as the sample is passed through the microfluidic passage, the at least one first voltage having first characteristics; and applying at least one second voltage to the at least one electrode as the second fluid is passed through the microfluidic passage after the sample is passed through the microfluidic passage, the at least one second voltage having second characteristics different than the first characteristics.

76. The method of claim 75, wherein the first characteristics include a first amplitude and a first frequency for the at least one first voltage, the second characteristics include a second amplitude and a second frequency for the at least one second voltage.

77. The method of 76, wherein the second amplitude is lower than the first amplitude.

78. The method of claim 76, wherein the second frequency is different than the first frequency.

79. The method of claim 71, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: subtracting the first image from the second image to generate a third image; and detecting microorganisms in the sample based on the third image.

80. The method of claim 71, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: subtracting the first image from the second image to generate a third image; and quantifying microorganisms in the sample based on the third image.

81. The method of claim 71, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: identifying in the first image, a first set of objects; identifying in the second image, a second set objects; removing the first set of objects from the second set of objects; and detecting microorganisms in the sample based on the second set of objects after the first set of objects has been removed.

82. The method of claim 71, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: quantifying in the first image, a number of first objects; quantifying in the second image, a number of second objects; and quantifying microorganisms in the sample based, at least in part, on the number of first objects and the number of second objects.

83. The method of claim 82, wherein quantifying in the first image, a number of first objects comprises applying a classifier to the first image to detect the first objects.

84. The method of claim 83, wherein quantifying, in the second image, a number of second objects comprises applying the classifier to the second image to detect the second objects.

85. The method of claim 71, wherein the stain is configured to selectively interact with the microorganisms in the sample.

86. The method of claim 71, wherein the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

87. A system for detecting and/or quantifying microorganisms in a sample, the sample including microorganisms and other components, the system comprising: a microfluidic passage for receiving the sample; a pump configured to pump a first fluid comprising a stain, a second fluid comprising a controlled solution without the stain, or the sample through the microfluidic passage; at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, particles of the sample on a surface of the at least one electrode using dielectrophoresis; an optical system configured to capture: a first image of the surface of the at least one electrode prior to pumping the sample through the microfluidic passage, and a second image of the surface of the at least one electrode after pumping the sample through the microfluidic passage; and at least one computing device configured to: detect and/or quantify microorganisms in the sample based on the first image and the second image.

88. The system of claim 87, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: subtracting the first image from the second image to generate a third image; and detecting microorganisms in the sample based on the third image.

89. The system of claim 87, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: subtracting the first image from the second image to generate a third image; and quantifying microorganisms in the sample based on the third image.

90. The system of claim 87, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: identifying in the first image, a first set of objects; identifying in the second image, a second set objects; removing the first set of objects from the second set of objects; and detecting microorganisms in the sample based on the second set of objects after the first set of objects has been removed.

91. The system of claim 87, wherein detecting and/or quantifying microorganisms in the sample based on the first image and the second image comprises: quantifying in the first image, a number of first objects; quantifying in the second image, a number of second objects; and quantifying microorganisms in the sample based, at least in part, on the number of first objects and the number of second objects.

92. The system of claim 91, wherein quantifying in the first image, a number of first objects comprises applying a classifier to the first image to detect the first objects.

93. The system of claim 92, wherein quantifying, in the second image, a number of second objects comprises applying the classifier to the second image to detect the second objects.

94. The system of claim 87, wherein the stain is configured to selectively interact with the microorganisms in the sample.

95. The system of claim 87, wherein the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

96. A method for improved detection and/or quantification of microorganisms in a sample, the sample including microorganisms and other components, the method comprising: passing a first fluid comprising a first stain through a microfluidic passage of a microfluidic device, the microfluidic passage including at least one electrode disposed therein; passing a second fluid through the microfluidic passage after passing the first fluid through the microfluidic passage, the second fluid comprising a controlled solution without the first stain; passing the sample through the microfluidic passage after passing the second fluid through the microfluidic passage, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the sample is passed through the microfluidic passage; passing a third fluid comprising a second stain through the microfluidic passage while the microorganisms remain captured on the surface of the at least one electrode; passing the second fluid through the microfluidic passage after passing the third fluid through the microfluidic passage and while the microorganisms remain captured on the surface of the at least one electrode; capturing, with an optical system, an image of the surface of the at least one electrode; and detecting and/or quantifying microorganisms in the sample based on the image.

97. The method of claim 96, further comprising: passing the second fluid through the microfluidic passage prior to passing the first fluid through the microfluidic passage.

98. The method of claim 96, further comprising: passing the second fluid through the microfluidic passage prior to passing the third fluid through the microfluidic passage.

99. The method of claim 96, wherein passing the first fluid, the second fluid, the third fluid, or the sample through the microfluidic passage comprises pumping the first fluid, the second fluid, the third fluid, or the sample through the microfluidic passage.

100. The method of claim 96, further comprising: applying at least one first voltage to the at least one electrode as the sample is passed through the microfluidic passage, the at least one first voltage having first characteristics; and applying at least one second voltage to the at least one electrode as the third fluid is passed through the microfluidic passage, the at least one second voltage having second characteristics different than the first characteristics.

101. The method of claim 96, wherein the first characteristics include a first amplitude and a first frequency for the at least one first voltage, the second characteristics include a second amplitude and a second frequency for the at least one second voltage.

102. The method of 101, wherein the second amplitude is lower than the first amplitude.

103. The method of claim 101, wherein the second frequency is different than the first frequency.

104. The method of claim 96, wherein detecting and/or quantifying microorganisms in the sample based on the image comprises: detecting microorganisms in the sample as objects in the image stained with the second stain but not the first stain.

105. The method of claim 104, wherein detecting and/or quantifying microorganisms in the sample based on the image further comprises: quantifying the detected microorganisms in the sample.

106. The method of claim 96, wherein the optical system includes a first filter and a second filter, and wherein capturing an image of the surface of the at least one electrode comprises capturing the image using the first filter and the second filter.

107. The method of claim 96, wherein the second stain is configured to selectively interact with the microorganisms in the sample.

108. The method of claim 107, wherein the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

109. A system for detecting and/or quantifying microorganisms in a sample, the sample including microorganisms and other components, the system comprising: a microfluidic passage for receiving the sample; a pump configured to pump a first fluid comprising a first stain, a second fluid comprising a controlled solution without the first stain, a third fluid comprising a third stain, or the sample through the microfluidic passage; at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, particles of the sample on a surface of the at least one electrode using dielectrophoresis; an optical system configured to capture an image of the surface of the at least one electrode after pumping the sample through the microfluidic passage; and at least one computing device configured to detect and/or quantify microorganisms in the sample based on the image.

110. The system of claim 109, wherein detecting and/or quantifying microorganisms in the sample based on the image comprises: detecting microorganisms in the sample as objects in the image stained with the second stain but not the first stain.

111. The system of claim 110, wherein detecting and / or quantifying microorganisms in the sample based on the image further comprises: quantifying the detected microorganisms in the sample.

112. The system of claim 109, wherein the optical system includes a first filter and a second filter, and wherein capturing an image of the surface of the at least one electrode comprises capturing the image using the first filter and the second filter.

113. The system of claim 109, wherein the second stain is configured to selectively interact with the microorganisms in the sample.

114. The method of claim 113, wherein the stain is metabolic stain configured to enter inside of the microorganisms in the sample.

114. A method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the other components including nucleic acids and/or organelles from degraded cells, the method comprising: treating the sample with an exonuclease configured to digest the nucleic acids to oligonucleotides less than five bases long to produce a clarified sample; passing the clarified sample through a microfluidic passage having at least one electrode disposed therein, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the clarified sample is passed through the microfluidic passage; capturing, with an optical system, at least one image of the surface of the at least one electrode; and quantifying microorganisms in the sample based on the at least one image.

116. The method of claim 115, wherein the sample is a sample from a bioreactor.

117. The method of claim 115, wherein capturing the at least one image comprises capturing a first image prior to passing the clarified sample through the microfluidic passage and capturing a second image after passing the clarified sample through the microfluidic passage, and quantifying microorganisms in the sample comprises quantifying microorganisms in the sample based on the first image and the second image.

118. The method of claim 115, further comprising passing a fluid comprising a stain through the microfluidic passage prior to capturing the at least one image.

119. A method for determining a quantity of microorganisms in a sample, the sample including microorganisms and other components, the other components including at least one surfactant, the method comprising: incubating the sample with activated carbon to adsorb the at least one surfactant in the sample to produce a clarified sample; passing the clarified sample through a microfluidic passage having at least one electrode disposed therein, wherein the microorganisms in the sample are captured on a surface of the at least one electrode using dielectrophoresis as the clarified sample is passed through the microfluidic passage; capturing, with an optical system, at least one image of the surface of the at least one electrode; and quantifying microorganisms in the sample based on the at least one image.

120. The method of claim 119, wherein the sample is a sample from a bioreactor.

121. The method of claim 119, wherein capturing the at least one image comprises capturing a first image prior to passing the clarified sample through the microfluidic passage and capturing a second image after passing the clarified sample through the microfluidic passage, and quantifying microorganisms in the sample comprises quantifying microorganisms in the sample based on the first image and the second image.

122. The method of claim 119, further comprising passing a fluid comprising a stain through the microfluidic passage prior to capturing the at least one image.

123. A method for identifying whether a sample is sterile, the method comprising: passing a first portion of the sample through a microfluidic passage having at least one electrode disposed therein; generating, with the at least one electrode, at least one dielectrophoretic force that acts on the first portion of the sample; determining, based on a response of the first portion of the sample to the at least one dielectrophoretic force that acts on the first portion of the sample, whether the first portion of the sample comprises one or more microorganisms; and when it is determined that the first portion of the sample does not comprise the one or more microorganisms, labeling the sample as sterile.

124. The method of claim 123, further comprising: providing an indication that the sample is sterile; and releasing the sample from sterility testing.

125. The method of claim 123, further comprising: when it is determined that the first portion of the sample does comprise the one or more microorganisms, labeling the sample as not sterile and providing an indication that a second portion of the sample is to be further processed.

126. The method of claim 123, wherein the one or more microorganisms comprise bacteria, mold, and/or yeast.

127. A system for identifying whether a sample is sterile, the system comprising: a microfluidic passage for receiving a first portion of the sample; at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, a plurality of particles of the first portion of the sample on a surface of the at least one electrode using dielectrophoresis; an optical system configured to capture at least one image of the surface of the at least one electrode; and at least one computing device configured to: determine, based on the at least one image, whether the first portion of the sample comprises one or more microorganisms; and when it is determined that the first portion of the sample does not comprise the one or more microorganisms, labeling the sample as sterile.

128. The system of claim 127, wherein the at least one computing device is further configured to: provide an indication that the sample is sterile; and release the sample from sterility testing.

129. The system of claim 128, wherein the at least one computing device is further configured to: when it is determined that the first portion of the sample does comprise the one or more microorganisms, label the sample as not sterile and provide an indication that a second portion of the sample is to be further processed.

130. The system of claim 127, wherein the one or more microorganisms comprise bacteria, mold, and/or yeast.

131. A method for identifying whether a sample is in a condition for release, the method comprising: passing a first portion of the sample into a microfluidic passage having at least one electrode disposed therein; generating, with the at least one electrode, at least one dielectrophoretic force that acts on the first portion of the sample; determining a number of microorganisms immobilized on a surface of the at least one electrode subsequent to the generating the at least one dielectrophoretic force; determining, based on the number of microorganisms immobilized on the surface of the at least one electrode, a concentration of microorganisms in the sample; and when it is determined that the concentration of microorganisms in the sample is less than a threshold concentration, labeling the sample as being in the condition for release.

132. A system for identifying whether a sample is in a condition for release, the system comprising: a microfluidic passage for receiving a first portion of the sample; at least one electrode disposed in the microfluidic passage, the at least one electrode configured to capture, when activated, a plurality of particles of the first portion of the sample on a surface of the at least one electrode using dielectrophoresis; an optical system configured to capture at least one image of the surface of the at least one electrode; and at least one computing device configured to: analyze the at least one image to determine a number of microorganisms immobilized on a surface of the at least one electrode subsequent to the generating the at least one dielectrophoretic force; determine whether the number of microorganisms immobilized on the surface of the at least one electrode is less than a threshold; and when it is determined that the number of microorganisms immobilized on the surface of the at least one electrode is less than the threshold, labeling the sample as being in the condition for release.