CN113870209A - Endoscope image identification system and equipment based on deep learning - Google Patents

Endoscope image identification system and equipment based on deep learning Download PDF

Info

Publication number
CN113870209A
CN113870209A CN202111108690.9A CN202111108690A CN113870209A CN 113870209 A CN113870209 A CN 113870209A CN 202111108690 A CN202111108690 A CN 202111108690A CN 113870209 A CN113870209 A CN 113870209A
Authority
CN
China
Prior art keywords
infection
deep learning
preset
model
image recognition
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.)
Pending
Application number
CN202111108690.9A
Other languages
Chinese (zh)
Inventor
于红刚
张梦娇
吴练练
邢达奇
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.)
Wuhan University WHU
Original Assignee
Wuhan University WHU
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 Wuhan University WHU filed Critical Wuhan University WHU
Priority to CN202111108690.9A priority Critical patent/CN113870209A/en
Publication of CN113870209A publication Critical patent/CN113870209A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • 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
    • 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
    • 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
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30092—Stomach; Gastric

Landscapes

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

Abstract

The invention discloses an endoscope image identification system and equipment based on deep learning, wherein the system is used for substituting an original gastroscope image into a preset convolution neural network model through a primary screening unit to obtain a qualified stomach image; the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images; the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic; the accuracy and effectiveness of helicobacter pylori infection judgment can be improved, a powerful diagnosis basis is provided for endoscopic physicians to diagnose HP infection, and meanwhile, endoscopic physicians can be assisted to analyze disease risks more effectively and more accurately.

Description

Endoscope image identification system and equipment based on deep learning
Technical Field
The invention relates to the technical field of medical auxiliary detection, in particular to an endoscope image recognition system and equipment based on deep learning.
Background
Helicobacter Pylori (HP) infection is one of the most common infections worldwide, with HP infection estimated to be 40% to 50% of the global population; research shows that any type of gastritis related to HP infection is at risk of developing gastric cancer, and the eradication of HP is an effective strategy for preventing gastric cancer; with the advancement of endoscopic technology, endoscopy has been used to diagnose gastritis, to determine the presence of HP infection and to assess the risk of gastric cancer.
HP infection is often nonspecific in endoscopic features and is distributed among multiple lesions and is not easily identifiable; according to the Kyoto Gastritis protocol issued in 2014 in Japan, the HP infection under endoscope is analyzed and summarized, so that the HP infection is easier to identify under endoscope and is beneficial to evaluating the risk of gastric cancer, and Zhao et al in the Accuracy of Endoscopic Diagnosis of Helicobacter pylori Based on the Kyoto Classification of gastric cancer, show that the Kyoto Gastritis type in Chinese population can be used for auxiliary Diagnosis of HP infection, and the existing way is to analyze and summarize the HP infection under endoscope, so that the HP infection is easier to identify under endoscope and is beneficial to evaluating the risk of gastric cancer; however, different characteristics suggest that the sensitivity, specificity and accuracy of HP infection have different effects, and a unified standard for determining whether HP infection exists is lacked.
Disclosure of Invention
The invention mainly aims to provide an endoscope image recognition system and equipment based on deep learning, and aims to solve the technical problems that in the prior art, HP infection is low in accuracy and effectiveness in judgment due to different characteristics of sensitivity, specificity and accuracy under an endoscope.
In a first aspect, the present invention provides an endoscopic image recognition system based on deep learning, which includes the following steps:
the primary screening unit is used for substituting the original gastroscope image into a preset convolutional neural network model to obtain a qualified stomach image;
the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images;
and the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic.
Optionally, the deep learning based endoscopic image recognition system further comprises: a model member unit for forming a model member,
the model construction unit is used for acquiring an input item data set of helicobacter pylori infection characteristics, and constructing a preset convolutional neural network model, the preset deep learning classification model and the infection characteristic identification model based on a residual error network according to the input item data set.
Optionally, the preliminary screening unit is further configured to construct a preset convolutional neural network model based on a residual error network, and substitute the original gastroscope image into the preset convolutional neural network model to obtain a qualified stomach image.
Optionally, the secondary screening unit is further configured to substitute the qualified stomach image into a preset deep learning classification model, filter out a normal stomach image, and obtain an abnormal image of the stomach.
Optionally, the identification unit is further configured to substitute the abnormal image into an infection feature identification model, obtain a feature matching a preset helicobacter pylori symptom set as an infection feature of helicobacter pylori, and determine a current infection state according to the infection feature.
Optionally, the identification unit is further configured to determine a current infection state according to a preset intelligent diagnosis formula and the infection characteristics.
Optionally, the identification unit is further configured to determine the number of infection features according to the infection features, and obtain an infection index through the following preset intelligent diagnosis formula;
Figure BDA0003273457410000021
wherein f is an infection index, exp is an exponential function, b is a regression coefficient, and xiThe number of the i-th relevant trait representing H.pylori infection, lambda being a regulatory parameter;
comparing the infection index with a preset infection threshold, and determining that the current infection state is an infection state when the infection index is greater than the preset infection threshold;
and when the infection index is not greater than the preset infection threshold value, determining that the current infection state is a non-infection state.
Optionally, the identification unit is further used for obtaining the correlation between each single characteristic in the infection characteristics and the helicobacter pylori infection;
the identification unit is further used for carrying out regression processing on the correlation of each single feature to obtain a regression coefficient of each single feature;
the identification unit is also used for fitting and cross-verifying each single characteristic through logistic regression, selecting an adjusting parameter with the minimum cross-verification error, and adding the adjusting parameter and the regression coefficient to obtain a logarithmic probability;
and obtaining a preset infection threshold value according to the logarithmic probability and a formula as follows:
Figure BDA0003273457410000031
wherein, the Probasic is a preset infection threshold, exp is an exponential function, and Log Odds is the logarithmic probability.
In a second aspect, to achieve the above object, the present invention further provides an endoscopic image recognition apparatus based on deep learning, wherein the endoscopic image recognition apparatus based on deep learning comprises: the system comprises a memory, a processor and a deep learning based endoscope image identification program stored on the memory and capable of running on the processor, wherein the deep learning based endoscope image identification program is configured to realize the functions of the deep learning based endoscope image identification system.
Optionally, the deep learning based endoscopic image recognition apparatus further comprises: an endoscope detector;
the endoscope detector is used for acquiring an original gastroscope map of a target user and feeding the original gastroscope map back to the processor.
The endoscope image identification system based on deep learning provided by the invention is used for substituting an original gastroscope image into a preset convolution neural network model through a primary screening unit to obtain a qualified stomach image; the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images; the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic; the accuracy and effectiveness of helicobacter pylori infection judgment can be improved, a powerful diagnosis basis is provided for endoscopic physicians to diagnose HP infection, and meanwhile, endoscopic physicians can be assisted to analyze disease risks more effectively and more accurately.
Drawings
FIG. 1 is a schematic diagram of an apparatus architecture of a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a schematic flowchart illustrating a first embodiment of an endoscopic image recognition system based on deep learning according to the present invention;
FIG. 3 is a flowchart illustrating a second embodiment of an endoscopic image recognition system based on deep learning according to the present invention;
fig. 4 is a convolutional neural network model training diagram in an endoscopic image recognition system based on deep learning.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The solution of the embodiment of the invention is mainly as follows: the primary screening unit is used for substituting the original gastroscope image into a preset convolutional neural network model to obtain a qualified stomach image; the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images; the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic; the accuracy and effectiveness of the judgment of the helicobacter pylori infection can be improved, a powerful diagnosis basis is provided for diagnosing the HP infection by an endoscope physician, the endoscope physician can be assisted to analyze the disease risk more effectively and more accurately, and the technical problems that in the prior art, the sensitivity, specificity and accuracy of different characteristics of the HP infection under the endoscope are different, and the accuracy and effectiveness of the judgment of the HP infection are lower are solved.
Referring to fig. 1, fig. 1 is a schematic device structure diagram of a hardware operating environment according to an embodiment of the present invention.
As shown in fig. 1, the apparatus may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Wherein a communication bus 1002 is used to enable connective communication between these components. The user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., a Wi-Fi interface). The Memory 1005 may be a high-speed RAM Memory or a Non-Volatile Memory (Non-Volatile Memory), such as a disk Memory. The memory 1005 may alternatively be a storage device separate from the processor 1001.
Those skilled in the art will appreciate that the configuration of the apparatus shown in fig. 1 is not intended to be limiting of the apparatus and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.
As shown in fig. 1, a memory 1005, which is a storage medium, may include therein an operating system, a network communication module, a user interface module, and an endoscopic image recognition program based on deep learning.
The apparatus of the present invention calls, through the processor 1001, an endoscopic image recognition program based on deep learning stored in the memory 1005, and performs the following operations:
the primary screening unit is used for substituting the original gastroscope image into a preset convolutional neural network model to obtain a qualified stomach image;
the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images;
and the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
the model construction unit is used for acquiring an input item data set of helicobacter pylori infection characteristics, and constructing a preset convolutional neural network model, the preset deep learning classification model and the infection characteristic identification model based on a residual error network according to the input item data set.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
the preliminary screening unit is further used for constructing a preset convolutional neural network model based on a residual error network, and substituting the original gastroscope image into the preset convolutional neural network model to obtain a qualified stomach image.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
and the secondary screening unit is also used for substituting the qualified stomach images into a preset deep learning classification model, filtering normal stomach images and obtaining abnormal images of the stomach.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
the identification unit is further used for substituting the abnormal image into an infection characteristic identification model, acquiring characteristics matched with a preset helicobacter pylori symptom set as infection characteristics of helicobacter pylori, and determining the current infection state according to the infection characteristics.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
the identification unit is further used for determining the current infection state according to a preset intelligent diagnosis formula and the infection characteristics.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
the identification unit is further used for determining the number of infection characteristics according to the infection characteristics and obtaining an infection index through the following preset intelligent diagnosis formula;
Figure BDA0003273457410000061
wherein f is an infection index, exp is an exponential function, b is a regression coefficient, and xiThe number of the i-th relevant trait representing H.pylori infection, lambda being a regulatory parameter;
comparing the infection index with a preset infection threshold, and determining that the current infection state is an infection state when the infection index is greater than the preset infection threshold;
and when the infection index is not greater than the preset infection threshold value, determining that the current infection state is a non-infection state.
The apparatus of the present invention calls, via the processor 1001, the deep learning based endoscope image recognition program stored in the memory 1005, and further performs the following operations:
the identification unit is further used for carrying out regression processing on the correlation of each single feature to obtain a regression coefficient of each single feature;
the identification unit is also used for fitting and cross-verifying each single characteristic through logistic regression, selecting an adjusting parameter with the minimum cross-verification error, and adding the adjusting parameter and the regression coefficient to obtain a logarithmic probability;
and obtaining a preset infection threshold value according to the logarithmic probability and a formula as follows:
Figure BDA0003273457410000071
wherein, the Probasic is a preset infection threshold, exp is an exponential function, and Log Odds is the logarithmic probability.
Further, the endoscope image recognition device based on deep learning further comprises: an endoscope detector;
the endoscope detector is used for acquiring an original gastroscope map of a target user and feeding the original gastroscope map back to the processor.
It can be understood that the endoscope detecting instrument is used for acquiring an original gastroscope image of a target user, and generally comprises a guide core and a probe, and the material of the endoscope detecting instrument can be soft, so that the probe is prevented from harming a target person to be detected, and the discomfort of the target person to be detected is reduced.
According to the scheme, the primary screening unit is used for substituting the original gastroscope image into the preset convolutional neural network model to obtain a qualified stomach image; the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images; the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic; the accuracy and effectiveness of helicobacter pylori infection judgment can be improved, a powerful diagnosis basis is provided for endoscopic physicians to diagnose HP infection, and meanwhile, endoscopic physicians can be assisted to analyze disease risks more effectively and more accurately.
Based on the hardware structure, the embodiment of the endoscope image identification system based on deep learning is provided.
Referring to fig. 2, fig. 2 is a flowchart illustrating a first embodiment of an endoscopic image recognition system based on deep learning according to the present invention.
In a first embodiment, the depth learning-based endoscopic image recognition system comprises the following steps:
and the primary screening unit 10 is used for substituting the original gastroscope image into a preset convolutional neural network model to obtain a qualified stomach image.
It should be noted that the original gastroscope image is an original detection image generated by a target person to be detected in a process of detecting the stomach, which is obtained through an endoscope detection device, and the preset convolutional neural network model is a convolutional neural network model which is preset and used for primarily screening and identifying the original stomach image.
And the secondary screening unit 20 is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images.
It can be understood that the preset deep learning classification model is a preset deep learning classification model for judging normal abnormal images in qualified stomach images, and abnormal images can be obtained by substituting the qualified stomach images into the preset deep learning classification model.
And the identification unit 30 is used for substituting the abnormal image into the infection characteristic identification model to obtain the infection characteristics of the helicobacter pylori, and determining the current infection state according to the infection characteristics.
It should be understood that the infection characteristic recognition model is a recognition model which is set in advance to recognize whether the infection characteristics of helicobacter pylori exist in the abnormal image, so that different current infection states are determined according to different infection characteristics.
According to the scheme, the primary screening unit is used for substituting the original gastroscope image into the preset convolutional neural network model to obtain a qualified stomach image; the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images; the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic; the accuracy and effectiveness of helicobacter pylori infection judgment can be improved, a powerful diagnosis basis is provided for endoscopic physicians to diagnose HP infection, and meanwhile, endoscopic physicians can be assisted to analyze disease risks more effectively and more accurately.
Further, fig. 3 is a schematic flow chart of a second embodiment of the endoscopic image recognition system based on deep learning according to the present invention, and as shown in fig. 3, the second embodiment of the endoscopic image recognition system based on deep learning according to the present invention is proposed based on the first embodiment, and in this embodiment, the endoscopic image recognition system based on deep learning further includes: a model member unit 40;
the model construction unit 40 is configured to obtain an input item data set of helicobacter pylori infection characteristics, and construct a preset convolutional neural network model, the preset deep learning classification model, and the infection characteristic identification model according to the input item data set based on a residual error network.
It should be noted that after the input item data set of the helicobacter pylori infection characteristics is obtained, several models, namely a preset convolutional neural network model, the preset deep learning classification model and the infection characteristic identification model, can be constructed based on the residual error network, so as to provide assistance for the identification of the endoscope image.
In a specific implementation, an intelligent diagnosis formula for judging helicobacter pylori HP infection in endoscopic images based on deep learning can be established:
Mhp＝f(X,Y,Z)
wherein M ishpAnd f is an intelligent diagnosis formula, X is an HP infection characteristic input item data set, Y is an HP infection parameter automatic optimization item data set, and Z is an HP infection comprehensive judgment item data set.
Accordingly, after the HP infection signature input data set is obtained, it may be based on a residual network such as: resnet50 builds deep learning models, namely a stomach image recognition model A (a preset convolutional neural network model), a positive abnormal image judgment model B (a preset deep learning classification model) and an HP infection characteristic recognition model C (an infection characteristic recognition model); generally, the model A \ B \ C needs to be trained and parameter adjusted, so that the model accuracy meets the requirement.
In a specific implementation, as shown in fig. 4, fig. 4 is a convolutional neural network model training diagram in an endoscope image recognition system based on deep learning, referring to fig. 4, after a gastroscope image is input into a sample database, an artificial mark can be performed according to picture attributes, the artificial mark is marked as esophagus, duodenum or other features, and HP infection related features, through an HP infection intelligent diagnosis formula, whether HP infection exists can be determined, meanwhile, the sample database and the gastroscope image can be subjected to machine training and learning, the accuracy rate is verified, a convolutional neural network model is generated when the accuracy rate passes, reasons are analyzed when the accuracy rate fails to pass, then the model is optimized, and the above steps are performed again until the accuracy rate passes; the generated preset convolutional neural network model can be used for identifying esophagus, duodenum and HP infection related characteristics (changes in chicken skin, atrophy, enteromorphia, macular tumors, mucosa swelling, plica snakelike swelling, white fluid turbidity, diffuse redness, punctate redness and hyperplastic polyps) and other characteristics.
According to the scheme, the preset convolutional neural network model, the preset deep learning classification model and the infection characteristic identification model are constructed on the basis of the residual error network according to the input item data set by obtaining the input item data set of the helicobacter pylori infection characteristics, so that the speed and the efficiency can be improved, and the accuracy and the effectiveness of judging the helicobacter pylori infection are further improved.
Further, with continuing reference to fig. 2, the preliminary screening unit in fig. 2 is further configured to construct a preset convolutional neural network model based on the residual error network, and substitute the original gastroscope image into the preset convolutional neural network model to obtain a qualified stomach image.
It can be understood that a preset convolutional neural network model can be constructed based on a residual error network, the original gastroscope image is substituted into the preset convolutional neural network model, namely the white light gastroscope original image to be classified is input, deep learning classification is carried out by adopting the preset convolutional neural network model, and esophagus, duodenum, fuzzy and qualified stomach images can be identified from the original gastroscope image.
Further, with reference to fig. 2, the secondary screening unit in fig. 2 is further configured to substitute the qualified stomach image into a preset deep learning classification model, filter out a normal stomach image, and obtain an abnormal stomach image.
It should be noted that, the qualified stomach images are substituted into the preset deep learning classification model to perform secondary screening, that is, normal images and abnormal images are identified from the qualified stomach images.
Further, with continuing reference to fig. 2, the identification unit in fig. 2 is further configured to substitute the abnormal image into an infection characteristic identification model, obtain a characteristic matching a preset helicobacter pylori symptom set as an infection characteristic of helicobacter pylori, and determine a current infection state according to the infection characteristic.
It should be understood that, by substituting the abnormal image into the infection characteristic identification model, it is possible to further identify characteristics matching a preset helicobacter pylori symptom set, and further use these characteristics as infection characteristics of helicobacter pylori, which may be chicken skin-like changes, atrophy, intestinal metaplasia, macular tumor, mucosal swelling, plica serpentine swelling, white fluid turbidity, diffuse redness, punctate redness, hyperplastic polyps and other HP infection-related characteristics, and this embodiment is not limited thereto; after the infection characteristics are obtained, different current infection states can be determined based on the different infection characteristics.
Further, with continued reference to fig. 2, the identification unit in fig. 2 is further configured to determine a current infection status according to a preset intelligent diagnosis formula and the infection characteristics.
It should be noted that the preset intelligent diagnosis formula is a preset formula of an infection index corresponding to infection characteristics of helicobacter pylori, and the current infection state can be determined through the infection characteristics and the preset intelligent diagnosis formula.
Further, with continued reference to fig. 2, the identification unit in fig. 2 is further configured to determine the number of infection features according to the infection features, and obtain an infection index through the following preset intelligent diagnosis formula;
Figure BDA0003273457410000101
wherein f is an infection index, exp is an exponential function, b is a regression coefficient, and xiThe number of the i-th relevant trait representing H.pylori infection, lambda being a regulatory parameter;
comparing the infection index with a preset infection threshold, and determining that the current infection state is an infection state when the infection index is greater than the preset infection threshold;
and when the infection index is not greater than the preset infection threshold value, determining that the current infection state is a non-infection state.
It should be understood that the preset infection threshold is a preset threshold for determining whether the current infection status is a certain infection status, and the current infection status can be quickly determined by comparing the infection index with the preset infection threshold.
In a specific implementation, the infection index f is greater than the preset infection threshold α, and HP infection is considered to be positive, i.e., the output result is positive, otherwise, the output result is negative.
Further, with continued reference to fig. 2, the identification unit of fig. 2 is further configured to obtain a correlation between individual ones of the infection signatures and helicobacter pylori infection;
the identification unit is further used for carrying out regression processing on the correlation of each single feature to obtain a regression coefficient of each single feature;
the identification unit is also used for fitting and cross-verifying each single characteristic through logistic regression, selecting an adjusting parameter with the minimum cross-verification error, and adding the adjusting parameter and the regression coefficient to obtain a logarithmic probability;
and obtaining a preset infection threshold value according to the logarithmic probability and a formula as follows:
Figure BDA0003273457410000111
wherein, the Probasic is a preset infection threshold, exp is an exponential function, and Log Odds is the logarithmic probability.
In specific implementation, the parameter of the HP infection model can be automatically optimized, namely, a deep learning classification model is adopted, and a single feature is used for judging whether the HP infection exists or not, so that the correlation (sensitivity, specificity, negative prediction value and positive prediction value) between each feature and the HP infection is obtained; based on the correlation between the single characteristics and the HP infection, regression processing can be carried out by using a 'glmnet' software package in R statistical software to obtain a regression coefficient of each characteristic; of course, the regression coefficient of each single feature may also be obtained in other manners, for example, the regression coefficient is obtained by way of model parameter adjustment, which is not limited in this embodiment;
it should be understood that, through the obtained regression coefficient, fitting different characteristics by using logistic regression to obtain a model, and meanwhile, in order to avoid overfitting of the model, adopting a cross validation method and selecting an adjustment parameter lambda value with the minimum cross validation error; further, the comprehensive judgment of H infection can be carried out, which comprises the following steps:
and (4) based on the obtained correlation coefficient of each characteristic and the adjustment parameter lambda value, adopting the model D, and refitting the model by using all the data for optimization to obtain a model E.
1) In the model E, the obtained regression coefficient and the adjustment parameter of each feature are added to calculate Log Odds, and the method is as follows:
Figure BDA0003273457410000112
and converted into probabilities.
2) And judging the sensitivity and specificity of the HP infection according to the acquired single characteristics to perform Receiver Operating Characteristic (ROC) curve analysis, selecting the optimal sensitivity and specificity, setting the threshold value alpha of the model E, and finally forming an HP infection intelligent diagnosis formula f.
It can be understood that the characteristics related to HP infection can be effectively identified, the correlation between different characteristics and diagnosis HP infection can be obtained, namely, the correlation between different characteristics and diagnosis HP infection is further verified, regression processing is carried out on the correlation to obtain a regression coefficient of each characteristic, a cross-validation method is adopted, a model with the minimum error is selected to obtain an adjustment parameter lambda, and an intelligent diagnosis formula for judging HP infection is finally formed.
According to the scheme, the comprehensiveness of helicobacter pylori infection judgment can be further improved, the accuracy and the effectiveness of helicobacter pylori infection judgment are improved, a powerful diagnosis basis is provided for an endoscopic physician to diagnose HP infection, and meanwhile, the endoscopic physician can be assisted to analyze the disease risk more effectively and more accurately.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system 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, article, or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims (10)

1. An endoscopic image recognition system based on deep learning, comprising:
the primary screening unit is used for substituting the original gastroscope image into a preset convolutional neural network model to obtain a qualified stomach image;
the secondary screening unit is used for substituting the qualified stomach images into a preset deep learning classification model to obtain abnormal images;
and the identification unit is used for substituting the abnormal image into an infection characteristic identification model to obtain the infection characteristic of the helicobacter pylori and determining the current infection state according to the infection characteristic.
2. The deep learning based endoscopic image recognition system according to claim 1, wherein said deep learning based endoscopic image recognition system further comprises: a model member unit for forming a model member,
the model construction unit is used for acquiring an input item data set of helicobacter pylori infection characteristics, and constructing a preset convolutional neural network model, the preset deep learning classification model and the infection characteristic identification model based on a residual error network according to the input item data set.
3. The deep learning-based endoscopic image recognition system according to claim 1, wherein the preliminary screening unit is further configured to construct a preset convolutional neural network model based on the residual network, and to substitute the original gastroscope image into the preset convolutional neural network model to obtain a qualified stomach image.
4. The deep learning-based endoscopic image recognition system of claim 3, wherein the secondary screening unit is further configured to substitute the qualified stomach image into a preset deep learning classification model to filter out a normal stomach image and obtain an abnormal stomach image.
5. The deep learning-based endoscopic image recognition system according to claim 1, wherein said recognition unit is further configured to substitute said abnormal image into an infection feature recognition model, obtain a feature matching a preset helicobacter pylori symptom set as an infection feature of helicobacter pylori, and determine a current infection status according to said infection feature.
6. The deep learning-based endoscopic image recognition system according to claim 5, wherein said recognition unit is further configured to determine a current infection status according to a preset intelligent diagnosis formula and said infection characteristics.
7. The deep learning-based endoscopic image recognition system according to claim 6, wherein said recognition unit is further configured to determine the number of infection features according to said infection features, and obtain an infection index according to said predetermined intelligent diagnosis formula;
Figure FDA0003273457400000021
wherein f is an infection index, exp is an exponential function, b is a regression coefficient, and xiThe number of the i-th relevant trait representing H.pylori infection, lambda being a regulatory parameter;
comparing the infection index with a preset infection threshold, and determining that the current infection state is an infection state when the infection index is greater than the preset infection threshold;
and when the infection index is not greater than the preset infection threshold value, determining that the current infection state is a non-infection state.
8. The deep learning-based endoscopic image recognition system according to claim 7, wherein said recognition unit is further configured to obtain the correlation between each single one of said infection features and H.pylori infection;
the identification unit is further used for carrying out regression processing on the correlation of each single feature to obtain a regression coefficient of each single feature;
the identification unit is also used for fitting and cross-verifying each single characteristic through logistic regression, selecting an adjusting parameter with the minimum cross-verification error, and adding the adjusting parameter and the regression coefficient to obtain a logarithmic probability;
and obtaining a preset infection threshold value according to the logarithmic probability and a formula as follows:
Figure FDA0003273457400000022
wherein, Probablity is a preset infection threshold, exp is an exponential function, and LogOdds is the logarithmic probability.
9. An endoscope image recognition apparatus based on deep learning, characterized in that the endoscope image recognition apparatus based on deep learning comprises: a memory, a processor, and a deep learning based endoscopic image recognition program stored on the memory and executable on the processor, the deep learning based endoscopic image recognition program configured to implement the functions of the deep learning based endoscopic image recognition system according to any one of claims 1 to 8.
10. The deep learning based endoscopic image recognition apparatus according to claim 9, wherein said deep learning based endoscopic image recognition apparatus further comprises: an endoscope detector;
the endoscope detector is used for acquiring an original gastroscope map of a target user and feeding the original gastroscope map back to the processor.
CN202111108690.9A 2021-09-22 2021-09-22 Endoscope image identification system and equipment based on deep learning Pending CN113870209A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111108690.9A CN113870209A (en) 2021-09-22 2021-09-22 Endoscope image identification system and equipment based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111108690.9A CN113870209A (en) 2021-09-22 2021-09-22 Endoscope image identification system and equipment based on deep learning

Publications (1)

Publication Number Publication Date
CN113870209A true CN113870209A (en) 2021-12-31

Family

ID=78993119

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111108690.9A Pending CN113870209A (en) 2021-09-22 2021-09-22 Endoscope image identification system and equipment based on deep learning

Country Status (1)

Country Link
CN (1) CN113870209A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114464316A (en) * 2022-04-11 2022-05-10 武汉大学 Stomach abnormal risk grade prediction method, device, terminal and readable storage medium
CN114511749A (en) * 2022-04-19 2022-05-17 武汉大学 Image processing method, image processing device, computer equipment and storage medium

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114464316A (en) * 2022-04-11 2022-05-10 武汉大学 Stomach abnormal risk grade prediction method, device, terminal and readable storage medium
CN114511749A (en) * 2022-04-19 2022-05-17 武汉大学 Image processing method, image processing device, computer equipment and storage medium
CN114511749B (en) * 2022-04-19 2022-06-28 武汉大学 Image processing method, image processing device, computer equipment and storage medium

Similar Documents

Publication Publication Date Title
EP3876190B1 (en) Endoscopic image processing method and system and computer device
CN109543719B (en) Cervical atypical lesion diagnosis model and device based on multi-modal attention model
CN113870209A (en) Endoscope image identification system and equipment based on deep learning
CN110600122A (en) Digestive tract image processing method and device and medical system
CN110473615B (en) Information processing method and device
JP7487418B2 (en) Identifying autofluorescence artifacts in multiplexed immunofluorescence images
CN114782760B (en) Stomach disease picture classification system based on multitask learning
CN114830107A (en) Image processing system, image processing apparatus, endoscope system, interface, and image processing method
JP2018054443A (en) Information processing device, information processing system and program
US11678869B2 (en) Biopsy support device, endoscope device, biopsy support method, and biopsy support program
CN116664929A (en) Laryngoscope image multi-attribute classification method based on multi-modal information fusion
CN117671573B (en) Helicobacter pylori infection state identification method and device based on gastroscope image
Yilmaz et al. Deep convolutional neural networks for onychomycosis detection using microscopic images with KOH examination
JP7346600B2 (en) Cervical cancer automatic diagnosis system
CN115171889B (en) Small sample gastric tumor diagnosis system
CN114667452A (en) Method for determining the concentration of an analyte in a body fluid
AU2020255759A1 (en) Skin disease analyzing program, skin disease analyzing method, skin disease analyzing device, and skin disease analyzing system
CN114998644B (en) Tumor diagnosis system, construction method thereof, terminal device and storage medium
JP6422142B1 (en) Image diagnostic system for fertilized egg, image diagnostic program for fertilized egg, and image diagnostic method for fertilized egg.
CN115965617A (en) Digestive tract endoscopy auxiliary system based on integrated learning
KR102293527B1 (en) System and method for interpreting ouput of neural network
KR102270669B1 (en) An image receiving device that calculates an image including a plurality of lesions using artificial intelligence
CN117853794A (en) Method, device, equipment and storage medium for multitasking classification of upper gastrointestinal endoscope images
Zhang et al. Abnormal region detection in gastroscopic images by combining classifiers on neighboring patches
US20230109857A1 (en) System and method for automatic diagnosis of middle ear diseases from an otoscopic image

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