CN117115156B - Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation - Google Patents

Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation Download PDF

Info

Publication number
CN117115156B
CN117115156B CN202311372109.3A CN202311372109A CN117115156B CN 117115156 B CN117115156 B CN 117115156B CN 202311372109 A CN202311372109 A CN 202311372109A CN 117115156 B CN117115156 B CN 117115156B
Authority
CN
China
Prior art keywords
nasopharyngeal carcinoma
image
deep learning
segmentation
learning model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202311372109.3A
Other languages
Chinese (zh)
Other versions
CN117115156A (en
Inventor
马淑颖
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
SECOND HOSPITAL OF TIANJIN MEDICAL UNIVERSITY
Original Assignee
SECOND HOSPITAL OF TIANJIN MEDICAL UNIVERSITY
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by SECOND HOSPITAL OF TIANJIN MEDICAL UNIVERSITY filed Critical SECOND HOSPITAL OF TIANJIN MEDICAL UNIVERSITY
Priority to CN202311372109.3A priority Critical patent/CN117115156B/en
Publication of CN117115156A publication Critical patent/CN117115156A/en
Application granted granted Critical
Publication of CN117115156B publication Critical patent/CN117115156B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G06N3/0455—Auto-encoder networks; Encoder-decoder networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20092—Interactive image processing based on input by user
    • G06T2207/20104—Interactive definition of region of interest [ROI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30096—Tumor; Lesion
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Public Health (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to the technical field of image processing, in particular to a nasopharyngeal carcinoma image processing method and system based on double-model segmentation, wherein when the nasopharyngeal carcinoma image is segmented, a CT image is firstly subjected to rough segmentation to obtain an interested region, then the interested region is amplified through a bilinear interpolation algorithm and then is input into a second deep learning model for segmentation to obtain a focus region of nasopharyngeal carcinoma; the precision of model segmentation is improved; according to the invention, the sample set is screened through the content of patient complaints in the electronic medical record, on one hand, samples with incorrect diagnosis conclusion records or low diagnosis conclusion accuracy are screened, on the other hand, some samples with difficult and complicated diseases such as mixed tumors, metastatic tumors and associated tumors are removed through the screening, so that as many standard nasopharyngeal carcinoma patient samples are reserved as possible, and then the samples are amplified, so that the deep learning model can learn the nasopharyngeal carcinoma characteristics more fully, and the segmentation accuracy is improved.

Description

Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation
Technical Field
The invention relates to the technical field of image processing, in particular to a nasopharyngeal carcinoma image processing method and system based on double-model segmentation.
Background
Along with the improvement of the living standard of people, the number of people suffering from cancer diseases is in a trend of rising year by year, the cancer has great influence on the life and health of people, and medical images have the irreplaceable effects of other evaluation means in the aspects of diagnosing cancer diseases, recording the cancer disease process, evaluating treatment response and the like, so that the medical images become important reference information for doctors to diagnose the cancer diseases.
The nasopharyngeal carcinoma focus area of CT image is marked by manual marking, on one hand, manual marking is time-consuming and labor-consuming, on the other hand, the manual marking is subject to great subjective impression of marking doctor, and inaccurate marking is easy to cause; in view of this current situation, more and more doctors and scholars consider labeling and segmenting nasopharyngeal carcinoma images by means of a deep learning model.
In the prior art, when a deep learning model is adopted to divide the nasopharyngeal carcinoma image, the accuracy of model training is generally improved by increasing the number of training sets, however, the nasopharyngeal carcinoma image has the characteristic of difficult acquisition, and the requirement of improving the accuracy cannot be well met by increasing the number of the training sets.
On the other hand, when a deep learning model is used to process a CT image, the segmentation of the medical image is generally realized through a complex model, or a scheme of performing coarse segmentation and then fine segmentation on the CT image exists, however, for a nasopharyngeal carcinoma focus, the size of the nasopharyngeal carcinoma focus is generally smaller, the focus belongs to small target recognition, and the simple coarse segmentation on the CT image is performed without processing, and then fine segmentation is performed, which causes that the size of the focus to be segmented in an input image is less affected on the segmentation precision.
In view of the above, there is a need for a nasopharyngeal carcinoma image processing method and system based on dual-model segmentation that improves segmentation accuracy.
Disclosure of Invention
The invention aims to provide a nasopharyngeal carcinoma image processing method and system based on double-model segmentation, which are used for solving the problem of low nasopharyngeal carcinoma image segmentation precision in the prior art.
In order to solve the technical problems, the invention specifically provides the following technical scheme: a nasopharyngeal carcinoma image processing method based on dual-model segmentation comprises the following steps:
s1: acquiring an electronic medical record of a patient with nasopharyngeal carcinoma as a sample set;
s2: screening the sample set to delete unqualified samples in the sample set so as to obtain a sample set for training a deep learning model;
s3: preprocessing CT images in samples in the sample set;
s4: training a first deep learning model by adopting each preprocessed CT image, wherein the first deep learning model is a rough segmentation model and is used for obtaining a region of interest of the CT image;
s5: image amplification is carried out on each region of interest through a bilinear interpolation algorithm, so that an amplified region of interest is obtained;
s6: training a second deep learning model by adopting the amplified region of interest; the second deep learning model is a fine segmentation model and is used for segmenting and obtaining nasopharyngeal carcinoma focus in the CT image;
s7: and (3) inputting the nasopharyngeal carcinoma image to be segmented into the first deep learning model after the preprocessing of the S3 to obtain a rough segmentation image, and then inputting the rough segmentation image into the second deep learning model to obtain a segmentation result of the nasopharyngeal carcinoma image to be segmented.
Preferably, in the step S1, each sample in the obtained sample set includes at least a patient symptom description, a CT image, and a diagnosis conclusion.
Preferably, the S2 specifically includes:
s2.1: carrying out statistical treatment on the patient complaint information of the sample set by adopting statistical software to obtain common symptoms of a nasopharyngeal carcinoma patients;
preferably, the statistical software is SPSS16.0;
preferably, the means for obtaining the common symptoms of a nasopharyngeal carcinoma patient are: ordering the occurrence frequency of certain symptoms in the sample set from large to small, and taking the first a symptoms as common symptoms of nasopharyngeal carcinoma patients;
preferably, a has a value of 5; the common symptoms of the a nasopharyngeal carcinoma patients are: EB virus antibody detection is positive, nasal bleeding, long-time nasal obstruction, cervical tumor and headache;
s2.2: deleting samples of the patient complaint information of the samples in the sample set, wherein the samples do not all contain common symptoms of the a nasopharyngeal carcinoma patients;
s2.3: and deleting samples with multiple cancer complications in the sample conclusion in the sample set, so as to obtain a sample set for training the deep learning model.
Preferably, the S3 includes:
s3.1, carrying out normalization operation on the CT image;
preferably, the normalization process specifically includes: normalizing the CT image by a Z-scanning method, setting the mean value of the image to be 0 and the variance to be 1, and eliminating the anisotropy of the CT image;
s3.2, performing data cleaning operation on the CT image;
the data cleansing includes: deleting all blank slices with zero pixel values, and cutting each CT image to a non-zero pixel value area;
s3.3, performing sample amplification operation on the CT image;
preferably, the CT images in the sample are amplified using random rotation, scaling, radiometric transformation, and cropping.
Preferably, the first deep learning model is a convolutional neural network model, and the region of interest is a region where a nasopharyngeal carcinoma focus is located and a part of background region;
preferably, the bilinear interpolation algorithm is specifically: in the amplifying process, a bilinear interpolation algorithm is adopted, newly added pixel values are calculated by weighted average from the pixel values of 4 adjacent pixels in a 2X 2 area around the pixel in the region of interest, and thus the amplified region of interest is obtained;
preferably, the second deep learning model is a U-NET network model; the U-NET network model comprises an encoder network and a decoder network, wherein the encoder comprises a convolution layer and a pooling layer, the convolution layer adopts a convolution kernel with the size of 3 multiplied by 3, a ReLU function is used as an activation function, the pooling layer adopts maximum pooling to reduce the resolution of the feature map of the nasopharyngeal carcinoma, and after each pooling operation, the number of the feature map of the nasopharyngeal carcinoma is increased by 1 time, but the resolution is reduced by 4 times;
the decoder network comprises a convolution layer and an up-sampling layer, the convolution kernel of the convolution layer is the same as that of the encoder, the up-sampling layer adopts bilinear interpolation to enlarge the resolution of the feature map of the nasopharyngeal carcinoma, the last layer of the decoder uses the convolution check feature map with the size of 1 multiplied by 1 to process, and the segmentation result is output through a Softmax function.
Preferably, the loss function is used for judging whether the second deep learning model is trained;
specifically, the loss function is:
wherein,representing the true value +_>Representing a predicted value of the second deep learning model, wherein n is the training frequency; loss (Low Density) MSE Values as a function of the injury;
according to another aspect of the present invention, there is provided a nasopharyngeal carcinoma image processing system based on dual model segmentation, using a nasopharyngeal carcinoma image processing method based on dual model segmentation as described above, comprising:
the sample set acquisition module is used for acquiring an electronic medical record of a patient with nasopharyngeal carcinoma as a sample set;
the sample set screening module is used for screening the sample set to delete unqualified samples in the sample set so as to obtain a sample set for training a deep learning model;
the preprocessing module is used for preprocessing CT images in the samples in the sample set;
the first deep learning model training module is used for training a first deep learning model by adopting each preprocessed CT image, wherein the first deep learning model is a rough segmentation model and is used for obtaining a region of interest of the CT image;
the interest region amplifying module is used for amplifying the images of each interest region through a bilinear interpolation algorithm so as to obtain amplified interest regions;
the second deep learning model training module is used for training a second deep learning model by adopting the amplified region of interest; the second deep learning model is a fine segmentation model and is used for segmenting and obtaining nasopharyngeal carcinoma focus in the CT image;
and the nasopharyngeal carcinoma image segmentation module is used for inputting the nasopharyngeal carcinoma image to be segmented into the first deep learning model after the preprocessing of the S3 to obtain a rough segmentation image, and then inputting the rough segmentation image into the second deep learning model to obtain a segmentation result of the nasopharyngeal carcinoma image to be segmented.
Compared with the prior art, the invention has the following beneficial effects:
when the nasopharyngeal carcinoma image is segmented, a first deep learning model is trained, a CT image is roughly segmented to obtain an interested region, the interested region is amplified through a bilinear interpolation algorithm, and the interested region is input into a second deep learning model for segmentation to obtain a focus region of the nasopharyngeal carcinoma; the method comprises the steps of amplifying a region of interest to convert a disease focus region from a small target to a large target, so that the precision of model segmentation is improved;
according to the invention, the sample set is screened through the content of patient complaints in the electronic medical record, on one hand, samples with incorrect diagnosis conclusion records or low diagnosis conclusion accuracy are screened, on the other hand, some samples with difficult and complicated diseases such as mixed tumors, metastatic tumors and associated tumors are removed through the screening, so that as many standard nasopharyngeal carcinoma patient samples are reserved as possible, and then the samples are amplified, so that the deep learning model can learn the nasopharyngeal carcinoma characteristics more fully, and the segmentation accuracy is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It will be apparent to those of ordinary skill in the art that the drawings in the following description are exemplary only and that other implementations can be obtained from the extensions of the drawings provided without inventive effort.
FIG. 1 is a flow chart of a nasopharyngeal carcinoma image processing method based on dual model segmentation provided by an embodiment of the present invention;
FIG. 2 is a flowchart of a method for screening the sample set according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a nasopharyngeal carcinoma image processing system based on dual model segmentation according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The concepts related to the present application will be described with reference to the accompanying drawings. It should be noted that the following descriptions of the concepts are only for making the content of the present application easier to understand, and do not represent a limitation on the protection scope of the present application; meanwhile, the embodiments and features in the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
In a first embodiment, as shown in fig. 1, the present invention provides a nasopharyngeal carcinoma image processing method based on dual model segmentation, including the following steps:
s1: and obtaining an electronic medical record of the patient with the diagnosis conclusion of nasopharyngeal carcinoma as a sample set.
In the embodiment, the electronic medical record with the diagnosis conclusion of nasopharyngeal carcinoma in the electronic medical record of the hospital from 2000 to 2022 is screened as a sample set for training a deep learning model; the number of samples taken was 1078.
Each electronic medical record in the acquired sample set at least comprises patient symptom description, CT images, diagnosis conclusions and the like.
S2: and screening the sample set to delete unqualified samples in the sample set, thereby obtaining the sample set for training the deep learning model.
In the prior art, the accuracy of model training is generally improved by increasing the number of samples, however, for medical samples, it is generally difficult to obtain a large amount of medical sample data due to the consideration of patient privacy and the like, so that improving the accuracy of model training by increasing the number of training samples is a difficult improvement direction.
Therefore, the accuracy of the sample is improved, the sample set of the S1 is screened according to the content of the patient complaint in the electronic medical record, on one hand, samples with incorrect diagnosis conclusion records or low diagnosis conclusion accuracy are screened, on the other hand, some samples with difficult and complicated diseases such as mixed tumor, metastatic tumor, associated tumor and the like are removed through the screening, and further, as many standard nasopharyngeal carcinoma patient samples as possible are reserved, so that the deep learning model can learn the characteristics of nasopharyngeal carcinoma more fully, and the segmentation accuracy is improved.
Specifically, as shown in fig. 2, S2 specifically includes:
s2.1: and carrying out statistical treatment on the patient complaint information of the sample set by adopting statistical software to obtain common symptoms of a nasopharyngeal carcinoma patients.
Specifically, the statistical software is SPSS16.0.
SPSS16.0 is a common statistical analysis software in the medical field, which can realize statistics and analysis of various information in electronic medical records and can produce various analysis forms.
Specifically, the common symptoms of nasopharyngeal carcinoma patients are obtained by: and (3) ordering the occurrence frequency of a certain symptom in the sample set from large to small, wherein the first a symptoms are common symptoms of nasopharyngeal carcinoma patients.
In this example, the frequency of occurrence of EB virus antibody detection positive in the sample set was 71.3%, the frequency of occurrence of epistaxis was 70.6%, the long-term nasal obstruction was 61.9%, the neck tumor was 61.2%, the headache was 54.6%, the tinnitus was 39.3%, the hearing was 28.2%, the vision blur was 3.4%, and so on.
In this embodiment, a has a value of 5; common symptoms for a patients with nasopharyngeal carcinoma are: EB virus antibody detection is positive, nasal bleeding, long-time nasal obstruction, cervical tumor and headache.
S2.2: patient complaint information for deleting samples in the sample set does not all contain samples of common symptoms of a nasopharyngeal carcinoma patients.
In fact, tumor biopsies and images have a better indication of nasopharyngeal carcinoma, and biopsies are known as tumor-confirmed gold fingers, for some patients, when a nasopharyngeal carcinoma is diagnosed, the patient tends to go to a more authoritative hospital or institution biopsy for confirmation, resulting in less biopsy data in some samples.
In addition, the image data is difficult to intelligently identify, and each image is intelligently identified in a preprocessing stage, so that the time for segmenting the whole nasopharyngeal carcinoma image is definitely greatly increased; therefore, the present embodiment selects to process the patient complaint information.
S2.3: samples with multiple cancer complications in the sample conclusion in the sample set are deleted.
Through the operation of S2.3, the samples of the mixed tumor and the associated tumor can be eliminated, so that only the sample with the diagnosis conclusion of nasopharyngeal carcinoma is reserved.
It is worth emphasizing that by the treatment of S2.1-S2.3, most of the difficult and complicated symptoms are eliminated, so that the sample of the more standard nasopharyngeal carcinoma patient is kept as much as possible, and a sample set for training the deep learning model is obtained.
S3: the CT images in the samples in the sample set are preprocessed.
Specifically, S3 includes:
and S3.1, carrying out normalization operation on the CT image.
In order to make the intensity value of the CT image more uniform and improve the accuracy of nasopharyngeal carcinoma focus boundary identification, normalization processing is carried out on the CT image.
Specifically, the normalization process is specifically: and normalizing the CT image by a Z-scanning method, setting the mean value of the image to be 0 and the variance to be 1, and eliminating the anisotropy of the CT image.
And S3.2, performing data cleaning operation on the CT image.
The data cleaning comprises the following steps: all blank slices with zero pixel values are deleted and each CT image is cropped to a non-zero pixel value area.
And S3.3, performing sample amplification operation on the CT image.
After the processing of S2 is performed on the sample set, the number of samples remaining is small, so that the samples need to be amplified, and in this embodiment, the CT images in the samples are amplified by using random rotation, scaling, radiation conversion and clipping, so as to increase the number of samples.
S4: and training a first deep learning model by adopting each preprocessed CT image, wherein the first deep learning model is a rough segmentation model and is used for obtaining a region of interest of the CT image.
Specifically, the first deep learning model is a convolutional neural network model, and the region of interest is a region where a nasopharyngeal carcinoma focus is located and a partial background region.
That is, the rough segmentation model is used to initially locate the position of the nasopharyngeal carcinoma patient.
S5: and (3) carrying out image amplification on each region of interest through a bilinear interpolation algorithm, so as to obtain the amplified region of interest.
Through the above-mentioned rough segmentation, in practice, most of the background area is already segmented out, and at this time, only the focus area of nasopharyngeal carcinoma and part of the background area remain; however, since the lesion area of nasopharyngeal carcinoma is generally smaller, it is difficult to identify by a general deep learning model, and a very complex deep learning model, for example, a form of adding a attentional mechanism and multi-model cascade, is required to be used to improve the segmentation accuracy to a certain extent.
In the embodiment, the small target is converted into the large target by performing image amplification processing on the region of interest, so that the deep learning model can refine the characteristics of more focus regions, and the model segmentation accuracy is improved; at this time, there is a problem that the magnification processing does not affect the resolution of the CT image; to solve this problem, the present embodiment adopts a bilinear interpolation algorithm to amplify the region of interest, so that the image magnification is close to lossless magnification.
The bilinear interpolation algorithm is specifically: in the amplifying process, a bilinear interpolation algorithm is adopted, and newly added pixel values are calculated by weighted average from pixel values of 4 adjacent pixels in a 2×2 area around the pixel in the region of interest, so that the amplified region of interest is obtained.
Experiments on a plurality of sample sets show that the accuracy of the region of interest is lost after the image amplification processing of S5, but the error is small after the comparison of the region of interest and the amplified region of interest, and the segmentation effect on the final nasopharyngeal carcinoma is less.
In the present embodiment, the magnification is 1.33, and it is emphasized that an excessive magnification cannot be set, because if the magnification is excessive, a certain influence is exerted on the resolution of the image, and thus the subsequent image segmentation is affected.
S6: training a second deep learning model by adopting the amplified region of interest; the second deep learning model is a fine segmentation model and is used for segmentation to obtain nasopharyngeal carcinoma focus in the CT image.
Specifically, the second deep learning model is a U-NET network model; the U-NET network model includes two parts, namely an encoder network and a decoder network, wherein the encoder includes a convolution layer and a pooling layer, in this embodiment, the convolution layer uses a convolution kernel with a size of 3×3, uses a ReLU function as an activation function, and the pooling layer uses maximum pooling to reduce the resolution of the feature map of the nasopharyngeal carcinoma, and after each pooling operation, the number of feature maps of the nasopharyngeal carcinoma is increased by 1 time, but the resolution is reduced by 4 times.
The decoder network comprises a convolution layer and an up-sampling layer, the convolution kernel of the convolution layer has the same size as the encoder, the up-sampling layer adopts bilinear interpolation to enlarge the resolution of the feature map of nasopharyngeal carcinoma, the last layer of the decoder uses the convolution check feature map with the size of 1 multiplied by 1 to process, so that the number of channels is equal to the number of categories, and finally the segmentation result is output through a Softmax function.
In this embodiment, the loss function is used to determine whether the second deep learning model has completed training.
Specifically, the loss function is:
wherein,representing the true value +_>Representing the predicted value of the second deep learning model, n being the number of training times, loss MSE Is the value of the damage function.
S7: and (3) preprocessing the nasopharyngeal carcinoma image to be segmented, inputting the preprocessed nasopharyngeal carcinoma image to the first deep learning model to obtain a rough segmentation image, and inputting the rough segmentation image to the second deep learning model to obtain a segmentation result of the nasopharyngeal carcinoma image to be segmented.
In a second embodiment, as shown in fig. 3, a nasopharyngeal carcinoma image processing system based on dual-model segmentation uses a nasopharyngeal carcinoma image processing method based on dual-model segmentation as described above, including:
the sample set acquisition module is used for acquiring an electronic medical record of a patient with nasopharyngeal carcinoma as a sample set.
And the sample set screening module is used for screening the sample set to delete unqualified samples in the sample set so as to obtain a sample set for training the deep learning model.
And the preprocessing module is used for preprocessing CT images in the samples in the sample set.
The first deep learning model training module is used for training a first deep learning model by adopting each preprocessed CT image, and the first deep learning model is a rough segmentation model and is used for obtaining a region of interest of the CT image.
The interest region amplifying module is used for amplifying the image of each interest region through a bilinear interpolation algorithm, so that the amplified interest region is obtained.
The second deep learning model training module is used for training a second deep learning model by adopting the amplified region of interest; the second deep learning model is a fine segmentation model and is used for segmentation to obtain nasopharyngeal carcinoma focus in the CT image.
And the nasopharyngeal carcinoma image segmentation module is used for inputting the nasopharyngeal carcinoma image to be segmented into the first deep learning model after the preprocessing of the S3 to obtain a rough segmentation image, and inputting the rough segmentation image into the second deep learning model to obtain a segmentation result of the nasopharyngeal carcinoma image to be segmented.
In a third embodiment, the present embodiment includes a computer readable storage medium having a data processing program stored thereon, the data processing program being executed by a processor to perform the nasopharyngeal carcinoma image processing method based on the dual model segmentation of the first embodiment.
It will be apparent to one of ordinary skill in the art that embodiments herein may be provided as a method, apparatus (device), or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Including but not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer, and the like. Furthermore, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.
The description herein is with reference to flowchart illustrations and/or block diagrams of methods, apparatus (devices) and computer program products according to embodiments herein. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present application. As used in the specification and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method or apparatus comprising such elements.
It should also be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," and the like indicate an orientation or a positional relationship based on that shown in the drawings, and are merely for convenience of description and simplification of the description, and do not indicate or imply that the apparatus or element in question must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present application. Unless specifically stated or limited otherwise, the terms "mounted," "connected," and the like are to be construed broadly and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the terms in this application will be understood by those of ordinary skill in the art in a specific context.
The above examples and/or embodiments are merely illustrative of preferred embodiments and/or implementations for implementing the technology of the present invention, and are not intended to limit the implementation of the technology of the present invention in any way, and any person skilled in the art should consider the technology or embodiments substantially the same as the present invention when making minor changes or modifications to other equivalent embodiments without departing from the scope of the technical means disclosed in the present invention.
Specific examples are set forth herein to illustrate the principles and embodiments of the present application, and the description of the examples above is only intended to assist in understanding the methods of the present application and their core ideas. The foregoing is merely a preferred embodiment of the present application, and it should be noted that, due to the limited text expressions, there is objectively no limit to the specific structure, and it will be apparent to those skilled in the art that numerous modifications, adaptations or variations can be made thereto and that the above-described features can be combined in a suitable manner without departing from the principles of the present application; such modifications, variations and combinations, or the direct application of the concepts and aspects of the invention in other applications without modification, are intended to be within the scope of this application.

Claims (6)

1. The nasopharyngeal carcinoma image processing method based on the dual-model segmentation is characterized by comprising the following steps of:
s1: acquiring an electronic medical record of a patient with nasopharyngeal carcinoma as a sample set;
s2: screening the sample set to delete unqualified samples in the sample set so as to obtain a sample set for training a deep learning model;
the step S2 specifically comprises the following steps:
s2.1: carrying out statistical treatment on the patient complaint information of the sample set by adopting statistical software to obtain common symptoms of a nasopharyngeal carcinoma patients; the statistical software is SPSS16.0, and the mode for obtaining the common symptoms of the nasopharyngeal carcinoma patient is as follows: the frequency of symptoms in the sample set is ranked from big to small, and the first a symptoms are common symptoms of nasopharyngeal carcinoma patients;
s2.2: deleting samples of the patient complaint information of the samples in the sample set, wherein the samples do not all contain common symptoms of the a nasopharyngeal carcinoma patients;
s2.3: deleting samples with multiple concurrent cancers in sample conclusions in the sample set, thereby obtaining a sample set for training a deep learning model;
s3: preprocessing CT images in each sample in the sample set for training the deep learning model;
s4: training a first deep learning model by adopting each preprocessed CT image, wherein the first deep learning model is a rough segmentation model and is used for obtaining a region of interest of the CT image;
s5: image amplification is carried out on each region of interest through a bilinear interpolation algorithm, so that an amplified region of interest is obtained; in the step S5, the bilinear interpolation algorithm specifically includes: in the amplifying process, a bilinear interpolation algorithm is adopted, newly added pixel values are calculated by weighted average from the pixel values of 4 adjacent pixels in a 2X 2 area around the pixel in the region of interest, and thus the amplified region of interest is obtained; the region of interest is amplified to convert the lesion region from a small target to a large target, so that the segmentation accuracy of the second deep learning model is improved;
s6: training a second deep learning model by adopting the amplified region of interest, wherein the second deep learning model is a fine segmentation model and is used for segmenting to obtain nasopharyngeal carcinoma focus in the CT image;
s7: and (3) inputting the nasopharyngeal carcinoma image to be segmented into the first deep learning model after the preprocessing of the S3 to obtain a rough segmentation image, and then inputting the rough segmentation image into the second deep learning model to obtain a segmentation result of the nasopharyngeal carcinoma image to be segmented.
2. The method for processing a dual-model segmentation-based nasopharyngeal carcinoma image according to claim 1, wherein in said S1, each sample in said obtained sample set includes at least a patient symptom description, a CT image, and a diagnosis conclusion.
3. The nasopharyngeal carcinoma image processing method based on dual-model segmentation as set forth in claim 1, wherein a has a value of 5; the common symptoms of the a nasopharyngeal carcinoma patients are: EB virus antibody detection is positive, nasal bleeding, long-time nasal obstruction, cervical tumor and headache.
4. The method for processing a nasopharyngeal carcinoma image based on dual-model segmentation according to claim 1, wherein in said S6, said second deep learning model is a U-NET network model; the U-NET network model comprises an encoder network and a decoder network, wherein the encoder comprises a convolution layer and a pooling layer, the convolution layer adopts a convolution kernel with the size of 3 multiplied by 3, a ReLU function is used as an activation function, the pooling layer adopts maximum pooling to reduce the resolution of the feature map of the nasopharyngeal carcinoma, and after each pooling operation, the number of the feature map of the nasopharyngeal carcinoma is increased by 1 time, and the resolution is reduced by 4 times;
the decoder network comprises a convolution layer and an up-sampling layer, the convolution kernel of the convolution layer is the same as that of the encoder, the up-sampling layer adopts bilinear interpolation to enlarge the resolution of the feature map of the nasopharyngeal carcinoma, the last layer of the decoder uses the convolution check feature map with the size of 1 multiplied by 1 to process, and the segmentation result is output through a Softmax function.
5. The method for processing a nasopharyngeal carcinoma image based on dual-model segmentation according to claim 4, wherein in said S6, a loss function is used to determine whether said second deep learning model is trained;
the loss function is:
wherein y is i The true value is represented by a value that is true,representing a predicted value of the second deep learning model, wherein n is the training frequency; loss (Low Density) MSE Is the value of the damage function.
6. A dual model segmentation-based nasopharyngeal carcinoma image processing system, wherein the dual model segmentation-based nasopharyngeal carcinoma image processing method of any one of claims 1-5 is used, comprising:
the sample set acquisition module is used for acquiring an electronic medical record of a patient with nasopharyngeal carcinoma as a sample set;
the sample set screening module is used for screening the sample set to delete unqualified samples in the sample set so as to obtain a sample set for training a deep learning model;
the preprocessing module is used for preprocessing CT images in the samples in the sample set;
the first deep learning model training module is used for training a first deep learning model by adopting each preprocessed CT image, wherein the first deep learning model is a rough segmentation model and is used for obtaining a region of interest of the CT image;
the interest region amplifying module is used for amplifying the images of each interest region through a bilinear interpolation algorithm so as to obtain amplified interest regions;
the second deep learning model training module is used for training a second deep learning model by adopting the amplified region of interest; the second deep learning model is a fine segmentation model and is used for segmenting and obtaining nasopharyngeal carcinoma focus in the CT image;
and the nasopharyngeal carcinoma image segmentation module is used for inputting the nasopharyngeal carcinoma image to be segmented into the first deep learning model after the preprocessing of the S3 to obtain a rough segmentation image, and then inputting the rough segmentation image into the second deep learning model to obtain a segmentation result of the nasopharyngeal carcinoma image to be segmented.
CN202311372109.3A 2023-10-23 2023-10-23 Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation Active CN117115156B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311372109.3A CN117115156B (en) 2023-10-23 2023-10-23 Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311372109.3A CN117115156B (en) 2023-10-23 2023-10-23 Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation

Publications (2)

Publication Number Publication Date
CN117115156A CN117115156A (en) 2023-11-24
CN117115156B true CN117115156B (en) 2024-01-05

Family

ID=88798738

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311372109.3A Active CN117115156B (en) 2023-10-23 2023-10-23 Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation

Country Status (1)

Country Link
CN (1) CN117115156B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107610129A (en) * 2017-08-14 2018-01-19 四川大学 A kind of multi-modal nasopharyngeal carcinima joint dividing method based on CNN
CN108257134A (en) * 2017-12-21 2018-07-06 深圳大学 Nasopharyngeal Carcinoma Lesions automatic division method and system based on deep learning
CN110956635A (en) * 2019-11-15 2020-04-03 上海联影智能医疗科技有限公司 Lung segment segmentation method, device, equipment and storage medium
CN116091522A (en) * 2023-02-07 2023-05-09 武汉大学 Medical image segmentation method, device, equipment and readable storage medium
CN116503607A (en) * 2023-06-28 2023-07-28 天津市中西医结合医院（天津市南开医院） CT image segmentation method and system based on deep learning
CN116580037A (en) * 2023-07-10 2023-08-11 天津医科大学第二医院 Nasopharyngeal carcinoma image segmentation method and system based on deep learning

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107610129A (en) * 2017-08-14 2018-01-19 四川大学 A kind of multi-modal nasopharyngeal carcinima joint dividing method based on CNN
CN108257134A (en) * 2017-12-21 2018-07-06 深圳大学 Nasopharyngeal Carcinoma Lesions automatic division method and system based on deep learning
CN110956635A (en) * 2019-11-15 2020-04-03 上海联影智能医疗科技有限公司 Lung segment segmentation method, device, equipment and storage medium
CN116091522A (en) * 2023-02-07 2023-05-09 武汉大学 Medical image segmentation method, device, equipment and readable storage medium
CN116503607A (en) * 2023-06-28 2023-07-28 天津市中西医结合医院（天津市南开医院） CT image segmentation method and system based on deep learning
CN116580037A (en) * 2023-07-10 2023-08-11 天津医科大学第二医院 Nasopharyngeal carcinoma image segmentation method and system based on deep learning

Also Published As

Publication number Publication date
CN117115156A (en) 2023-11-24

Similar Documents

Publication Publication Date Title
CN108665456B (en) Method and system for real-time marking of breast ultrasound lesion region based on artificial intelligence
CN108898595B (en) Construction method and application of positioning model of focus region in chest image
Hansen et al. Ultrasonography of the kidney: a pictorial review
CN111429473B (en) Chest film lung field segmentation model establishment and segmentation method based on multi-scale feature fusion
CN111340827A (en) Lung CT image data processing and analyzing method and system
CN113538435B (en) Pancreatic cancer pathological image classification method and system based on deep learning
CN113420826A (en) Liver focus image processing system and image processing method
Langarizadeh et al. Improvement of digital mammogram images using histogram equalization, histogram stretching and median filter
CN115147404B (en) Intracranial aneurysm segmentation method by fusing dual features with MRA image
US20230162353A1 (en) Multistream fusion encoder for prostate lesion segmentation and classification
US20190035076A1 (en) Automated percentage of breast density measurements for full field digital mammography
CN114972266A (en) Lymphoma ultrasonic image semantic segmentation method based on self-attention mechanism and stable learning
Pociask et al. Differential diagnosis of cysts and granulomas supported by texture analysis of intraoral radiographs
CN117495882B (en) Liver tumor CT image segmentation method based on AGCH-Net and multi-scale fusion
CN117115156B (en) Nasopharyngeal carcinoma image processing method and system based on dual-model segmentation
CN112200810B (en) Multi-modal automated ventricle segmentation system and method of use thereof
CN117934499A (en) Kidney tumor ultrasonic image segmentation method based on improved UNet++, and method for segmenting kidney tumor ultrasonic image based on improved UNet++
CN117036343A (en) FFOCT image analysis method and device for identifying axillary lymph node metastasis
CN114937022A (en) Novel coronary pneumonia disease detection and segmentation method
Swami et al. An algorithmic detection of brain tumour using image filtering and segmentation of various radiographs
Nour et al. Skin lesion segmentation based on edge attention vnet with balanced focal tversky loss
Tummala et al. Curriculum learning based overcomplete U-Net for liver tumor segmentation from computed tomography images
Shaik Amir et al. CT brain image advancement for ICH diagnosis
CN116152502B (en) Medical image segmentation method and system based on decoding layer loss recall
CN116862872A (en) Mammary gland focus ultrasonic image processing model training method, related method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant