Artificial intelligence pancreatic pathology auxiliary diagnosis system

Abstract

The invention provides an artificial intelligence pancreas pathology auxiliary diagnosis system. The system comprises a pathology acquisition module which acquires pancreas pathology sections to obtain pancreas pathology sections; the data labeling module performs data labeling on the pancreatic pathological section to obtain a labeled pancreatic pathological section; the supervised learning module scans the pancreas pathological section with the label, and performs information enhancement and segmentation processing on a scanned image to obtain an image segmentation result; and the analysis processing module is used for performing feature analysis and prediction on the image segmentation result by adopting a distributed micro-processing architecture to obtain pancreatic pathology auxiliary diagnosis data. According to the method, intelligent processing and analysis of the pancreatic pathological section are realized, so that the pancreatic pathological section under any clinical manifestation condition can be identified and analyzed, artificial subjective analysis and judgment are not involved, the accuracy of pancreatic pathological auxiliary diagnosis data is improved, and the accuracy of the pancreatic pathological auxiliary diagnosis data is improved. And a pathologist can directly diagnose according to the pancreas pathology auxiliary diagnosis data, so that the burden of the pathologist is effectively relieved.

Description

Technical Field

[0001] The present invention relates to the field of biomedical computer information processing technology, and in particular to an artificial intelligence pancreatic pathology auxiliary diagnosis system. Background Art

[0003] In clinical medical practice, histopathological diagnosis is recognized as the gold standard for tumor diagnosis and can provide essential reference information for subsequent treatment. It plays a very important role. However, for the same type of tumor, different tissue regions, different stages of the disease, different environmental influences, different physical factors, etc., will also cause the tissue cell morphology in the pathological sections to vary greatly. Only when pathologists have accumulated a lot of experience can they make more accurate judgments on tissue sections. The differences in diagnostic experience of different pathologists and even the fluctuations in the reading status of the same pathologist will affect the accuracy of the final diagnosis. In addition, the clinical manifestations of chronic pancreatitis and pancreatic cancer are very similar to the manifestations of pathological sections, which are very easy to misdiagnose, affecting patients' treatment decisions and prognosis assessments. As the incidence of tumor diseases in the population is increasing and the complexity of tissue pathology diagnosis is increasing, it is difficult to meet the growing demand of patients for pathology diagnosis by relying on traditional methods. Therefore, the present invention proposes an artificial intelligence pancreatic pathology auxiliary diagnosis system to realize intelligent processing and analysis of pancreatic pathology sections, so that pancreatic pathology sections under any clinical manifestation can be identified and analyzed without involving human subjective analysis and judgment, thereby improving the accuracy of pancreatic pathology auxiliary diagnosis data and enabling pathologists to make diagnoses directly based on pancreatic pathology auxiliary diagnosis data, effectively reducing the burden on pathologists. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence pancreatic pathology auxiliary diagnosis system to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an artificial intelligence pancreatic pathology auxiliary diagnosis system, comprising: a pathology acquisition module, a data annotation module, a supervised learning module and an analysis and processing module;

[0006] The pathology acquisition module is used to collect and acquire pancreatic pathology slices to obtain pancreatic pathology slices;

[0007] The data annotation module is used to perform data annotation on the pancreatic pathological sections to obtain annotated pancreatic pathological sections;

[0008] The supervised learning module is used to scan the annotated pancreatic pathological sections and perform information enhancement and segmentation processing on the scanned images to obtain image segmentation results;

[0009] The analysis and processing module is used to perform feature analysis and prediction on the image segmentation results using a distributed microprocessing architecture to obtain pancreatic pathology auxiliary diagnosis data.

[0010] Preferably, the pathology acquisition module comprises: an acquisition unit, a preliminary judgment unit and an image preprocessing unit;

[0011] The acquisition and processing unit is used to acquire pancreatic pathological sections to obtain pancreatic pathological sections;

[0012] The preliminary judgment unit is used to perform preliminary analysis and judgment on the image quality of the pancreatic pathological slices to obtain preliminary image analysis and judgment results;

[0013] The image preprocessing unit is used to optimize the pancreatic pathological sections according to the preliminary analysis and judgment results of the images.

[0014] Preferably, the data annotation module comprises: a first annotation unit, a second annotation unit and a screening and review unit;

[0015] The first labeling unit is used to perform a first labeling on the pancreatic pathological section to obtain a first labeled pancreatic pathological section;

[0016] The second labeling unit performs a second labeling on the first labeled pancreatic pathology section to obtain a second labeled pancreatic pathology section;

[0017] The screening and review unit is used to compare and analyze the first annotation and the second annotation in the second annotated pancreatic pathological section, determine whether the first annotation and the second annotation are consistent, obtain analysis results, and screen out pancreatic pathological sections with inconsistent first and second annotations according to the analysis results for further review.

[0018] Preferably, the supervised learning module uses a weakly supervised learning framework to perform refined processing on the annotated pancreatic pathological sections, and the weakly supervised learning framework includes: a first processing unit and a second processing unit;

[0019] The first processing unit is used to scan the annotated pancreatic pathological sections, and use a combined multi-instance learning method to train and annotate the scanned images to obtain pixel-level annotations of image-level data;

[0020] The second processing unit is used to perform image segmentation model training and reasoning based on pixel-level annotation of image-level data to obtain image segmentation results of pancreatic pathological sections.

[0021] Preferably, the first processing unit scans annotated pancreatic pathological sections and uses a combined multi-instance learning method to train and annotate the scanned images, including:

[0022] Fully scan the annotated pancreatic pathology sections to determine image-level data and image-level annotations;

[0023] The combined multi-instance learning method is used to divide the image-level data into multiple image blocks according to the scale factor;

[0024] Sampling all image blocks in the fully scanned annotated pancreatic pathology slice as an image package, and performing judgment and prediction on the image package through a first deep learning module to obtain a first prediction result, and performing judgment and prediction on the image package through a second deep learning module to obtain a second prediction result;

[0025] Eliminate image blocks with different results according to the first prediction result and the second prediction result to obtain a representative image block;

[0026] By training the image block classification model on the representative image block, an image block classification optimization model is obtained;

[0027] The image block classification optimization model is used to perform inference prediction on the image blocks in the image-level data to obtain the prediction results of the hotspot areas of the pancreatic pathological sections;

[0028] Based on the prediction results of hotspot areas in pancreatic pathological sections, the image-level annotations are enhanced to the image block level to obtain the supervision information of the image blocks and obtain the image block-level annotations.

[0029] Preferably, after obtaining the image block-level annotation, the first processing unit performs pixel-level annotation on the image-level data and the image block-level annotation and then outputs them to the second processing unit; wherein, when the image-level data and the image block-level annotation are performed pixel-level annotation, the image block-level annotation is directly assigned to each pixel point.

[0030] Preferably, the second processing unit performs image segmentation model training and reasoning based on pixel-level annotation of image-level data, including:

[0031] Receiving image-level data and pixel-level annotations input by the first processing unit;

[0032] The image segmentation model is trained using image-level data and pixel-level annotations to obtain a trained image segmentation model;

[0033] The image-level data is input into the trained image segmentation model for image segmentation processing to obtain the image segmentation results of pancreatic pathological sections.

[0034] Preferably, when the analysis and processing module adopts a distributed microprocessing architecture to perform feature analysis and prediction on the image segmentation result, it includes:

[0035] A message queue is formed for the pancreatic pathological slice segmentation block in the image segmentation result;

[0036] Performing feature analysis and prediction on the pancreatic pathological slice segmentation block in the working node in turn through the microprocessor according to the message queue to obtain analysis and prediction data of the pancreatic pathological slice segmentation block, including: exploring feature information of the pancreatic pathological slice segmentation block in the image segmentation result from multiple angles to obtain auxiliary diagnosis data of the pancreatic pathological slice segmentation block, and performing prediction based on the auxiliary diagnosis data of the pancreatic pathological slice segmentation block to obtain prediction diagnosis data of the pancreatic pathological slice segmentation block;

[0037] For the analysis and prediction data of the pancreatic pathology section segmentation block, the auxiliary diagnosis data of the pancreatic pathology section segmentation block and the prediction diagnosis data of the pancreatic pathology section segmentation block are integrated together to obtain the pancreatic pathology auxiliary diagnosis data.

[0038] Preferably, there are multiple working nodes, and one working node corresponds to one microprocessor. When the microprocessors perform feature analysis and prediction on the pancreatic pathological slice segmentation blocks in the working nodes in sequence according to the message queue, multiple working nodes perform feature analysis and prediction in parallel.

[0039] Preferably, a scheduler is provided in the distributed microprocessor architecture, and the scheduler is used to allocate and manage tasks for working nodes according to a message queue.

[0040] The present invention realizes intelligent processing and analysis of pancreatic pathology sections through an artificial intelligence pancreatic pathology auxiliary diagnosis system, so that pancreatic pathology sections under any clinical manifestation situation can be accurately and efficiently identified, and the accuracy of pancreatic pathology auxiliary diagnosis data is improved, so that pathologists can directly make diagnoses based on the pancreatic pathology auxiliary diagnosis data, which can not only reduce the pressure of pathologists, but also improve the accuracy of pathologists' diagnosis. In addition, in the process of obtaining pancreatic pathology auxiliary diagnosis data, human subjective analysis and judgment are not involved, which reduces the influence of human subjective factors and ensures the objective accuracy of pancreatic pathology auxiliary diagnosis data. In addition, the data annotation module can be used to annotate key areas in pancreatic pathology sections, which facilitates the processing and analysis of pancreatic pathology sections, allowing the focus to be placed on the annotated parts and improving the recognition accuracy. The supervised learning module enables refined processing of annotated pancreatic pathology sections, improving the accuracy of annotations, and also provides an analyzed basis for the analysis and processing module. The analysis and processing module uses a distributed microprocessor architecture to perform parallel processing on image segmentation results, improving the efficiency of feature analysis and prediction, so that pancreatic pathology auxiliary diagnosis data can be obtained in a shorter time. Feature analysis and prediction can also be used to realize intelligent automatic analysis of pancreatic pathology sections in image segmentation results, which can not only extract characteristic information from pancreatic pathology sections and provide diagnostic support for pathologists, but also provide convenience for pathologists to make diagnoses and improve the accuracy of pathologists' diagnoses.

[0041] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0042] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0044] Figure 1 This is a schematic diagram of an artificial intelligence pancreatic pathology auxiliary diagnosis system according to the present invention;

[0045] Figure 2 It is a schematic diagram of a pathology acquisition module in an artificial intelligence pancreatic pathology auxiliary diagnosis system according to the present invention;

[0046] Figure 3 This is a schematic diagram of a data annotation module in an artificial intelligence pancreatic pathology auxiliary diagnosis system according to the present invention;

[0047] Figure 4 It is a schematic diagram of a weakly supervised learning framework of a supervised learning module in an artificial intelligence pancreatic pathology auxiliary diagnosis system according to the present invention;

[0048] Figure 5 It is a schematic diagram of a first processing unit of a weakly supervised learning framework in an artificial intelligence pancreatic pathology auxiliary diagnosis system according to the present invention;

[0049] Figure 6 It is a schematic diagram of a second processing unit of a weakly supervised learning framework in an artificial intelligence pancreatic pathology auxiliary diagnosis system according to the present invention;

[0050] Figure 7 This is a schematic diagram of the steps of the analysis and processing module in the artificial intelligence pancreatic pathology auxiliary diagnosis system described in the present invention. DETAILED DESCRIPTION

[0051] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0052] The embodiment of the present invention provides an artificial intelligence pancreatic pathology auxiliary diagnosis system, including: a pathology acquisition module, a data annotation module, a supervised learning module and an analysis and processing module;

[0053] The pathology acquisition module is used to collect and acquire pancreatic pathology slices to obtain pancreatic pathology slices;

[0054] The data annotation module is used to perform data annotation on the pancreatic pathological sections to obtain annotated pancreatic pathological sections;

[0055] The supervised learning module is used to scan the annotated pancreatic pathological sections and perform information enhancement and segmentation processing on the scanned images to obtain image segmentation results;

[0056] The analysis and processing module is used to perform feature analysis and prediction on the image segmentation results using a distributed microprocessing architecture to obtain pancreatic pathology auxiliary diagnosis data.

[0057] In the above technical solution, the pancreatic pathological slices acquired by the pathology acquisition module are digital slice images.

[0058] In the above technical solution, if Figure 1 As shown, the pathology acquisition module, the data annotation module, the supervised learning module and the analysis and processing module are connected in sequence.

[0059] In the above technical solution, the supervised learning module and the analysis and processing module use artificial intelligence to automatically scan annotated pancreatic pathology sections, perform information enhancement and segmentation processing on the scanned images, and use a distributed microprocessing architecture to perform feature analysis and prediction on the image segmentation results.

[0060] The above technical solution realizes intelligent processing and analysis of pancreatic pathology sections through the artificial intelligence pancreatic pathology auxiliary diagnosis system, so that pancreatic pathology sections under any clinical manifestation situation can be accurately and efficiently identified, thereby improving the accuracy of pancreatic pathology auxiliary diagnosis data, so that pathologists can make diagnoses directly based on pancreatic pathology auxiliary diagnosis data, which can not only reduce the pressure of pathologists, but also improve the accuracy of pathologists' diagnosis. In addition, human subjective analysis and judgment are not involved in the process of obtaining pancreatic pathology auxiliary diagnosis data, which reduces the influence of human subjective factors and ensures the objective accuracy of pancreatic pathology auxiliary diagnosis data. In addition, the data annotation module can be used to annotate key areas in pancreatic pathology sections, which facilitates the processing and analysis of pancreatic pathology sections, allowing the focus to be placed on the annotated parts and improving the recognition accuracy. The supervised learning module enables refined processing of annotated pancreatic pathology sections, improving the accuracy of annotations, and also provides an analyzed basis for the analysis and processing module. The analysis and processing module uses a distributed microprocessor architecture to perform parallel processing on image segmentation results, improving the efficiency of feature analysis and prediction, so that pancreatic pathology auxiliary diagnosis data can be obtained in a shorter time. Feature analysis and prediction can also be used to realize intelligent automatic analysis of pancreatic pathology sections in image segmentation results, which can not only extract characteristic information from pancreatic pathology sections and provide diagnostic support for pathologists, but also provide convenience for pathologists to make diagnoses and improve the accuracy of pathologists' diagnoses.

[0061] like Figure 2 As shown, in one embodiment provided by the present invention, the pathology acquisition module includes: an acquisition unit, a preliminary judgment unit and an image preprocessing unit;

[0062] The acquisition and processing unit is used to acquire pancreatic pathological sections to obtain pancreatic pathological sections;

[0063] The preliminary judgment unit is used to perform preliminary analysis and judgment on the image quality of the pancreatic pathological slices to obtain preliminary image analysis and judgment results;

[0064] The image preprocessing unit is used to optimize the pancreatic pathological sections according to the preliminary analysis and judgment results of the images.

[0065] In the above technical solution, the preliminary judgment unit performs pixel analysis on the pancreatic pathological section, determines the image resolution and clarity of the pancreatic pathological section, and performs standardized judgment based on the image resolution and clarity, thereby obtaining a preliminary analysis and judgment result of the image, wherein the judgment criteria involved in the standardized judgment are pre-set.

[0066] In the above technical solution, when the preliminary judgment unit performs pixel analysis on the pancreatic pathological section, it performs preliminary identification on the pancreatic pathological section, locks the area where the object image exists in the pancreatic pathological section, and divides the pancreatic pathological section into a first preliminary identification area and a second preliminary identification area, so that after determining the image resolution and clarity of the pancreatic pathological section, when performing standardized judgment based on the image resolution and clarity, the first judgment standard is used to perform standardized judgment on the first preliminary identification area, and the second judgment standard is used to perform standardized judgment on the second preliminary identification area, so as to obtain a preliminary image analysis judgment result, wherein the first preliminary identification area is the area where the object image exists in the pancreatic pathological section, and the second preliminary identification area is the remaining area in the pancreatic pathological section except the object image, and the first judgment standard has higher requirements on image resolution and clarity than the second judgment standard.

[0067] In the above technical solution, when the image preprocessing unit optimizes the pancreatic pathological section according to the preliminary image analysis result, when the preliminary image analysis result determines that the pancreatic pathological section does not meet the standard, the pancreatic pathological section is optimized.

[0068] The above technical solution optimizes pancreatic pathological sections with low image quality through the preliminary judgment unit and the image preprocessing unit, thereby ensuring the pancreatic pathological sections, so that the data annotation module can more clearly understand the contents of the pancreatic pathological sections when processing the pancreatic pathological sections, thereby improving the annotation accuracy of the data annotation module.

[0069] like Figure 3 As shown, in one embodiment provided by the present invention, the data annotation module includes: a first annotation unit, a second annotation unit and a screening and review unit;

[0070] The first labeling unit is used to perform a first labeling on the pancreatic pathological section to obtain a first labeled pancreatic pathological section;

[0071] The second labeling unit performs a second labeling on the first labeled pancreatic pathology section to obtain a second labeled pancreatic pathology section;

[0072] The screening and review unit is used to compare and analyze the first annotation and the second annotation in the second annotated pancreatic pathological section, determine whether the first annotation and the second annotation are consistent, obtain analysis results, and screen out pancreatic pathological sections with inconsistent first and second annotations according to the analysis results for further review.

[0073] In the above technical solution, the first marking unit and the second marking unit perform marking based on the marking system of iPad and Pencil.

[0074] In the above technical solution, the first labeling unit and the second labeling unit are different objects that label the pancreatic pathological slices.

[0075] In the above technical solution, after the first marking is completed, the first marking unit is allocated to the second marking unit for second marking.

[0076] In the above technical solution, the second annotated pancreatic pathological section includes both the result of the first annotation and the result of the second annotation.

[0077] In the above technical solution, when the pancreatic pathological sections with inconsistent first and second annotations are screened out for further review according to the analysis results, the pancreatic pathological sections with inconsistent first and second annotations are submitted to the superior leader as questionable data, and the superior leader reviews the questionable data.

[0078] In the above technical solution, the pancreatic pathological sections with inconsistent first and second annotations are scanned by the supervised learning module after being reviewed and approved, otherwise the supervised learning module does not scan the pancreatic pathological sections with inconsistent first and second annotations.

[0079] The above technical solution uses the first and second marking units to perform the first and second markings respectively, so that the two markings are compared, and self-correction is achieved, thereby reducing the marking error. In addition, the second marking unit performs the second standard based on the first marked pancreatic pathological section obtained by the first marking unit, which can facilitate the screening and review module to perform comparative analysis of the first and second markings, making it easier to determine whether the first and second markings are consistent, thereby improving the efficiency of the screening and review unit. In addition, the iPad and Pencil marking system is easy to use, which can greatly improve the efficiency of the first and second markings performed by the first and second marking units.

[0080] In one embodiment provided by the present invention, the supervised learning module uses a weakly supervised learning framework to perform fine processing on annotated pancreatic pathological sections, such as Figure 4 As shown, the weakly supervised learning framework includes: a first processing unit and a second processing unit;

[0081] The first processing unit is used to scan the annotated pancreatic pathological sections, and use a combined multi-instance learning method to train and annotate the scanned images to obtain pixel-level annotations of image-level data;

[0082] The second processing unit is used to perform image segmentation model training and reasoning based on pixel-level annotation of image-level data to obtain image segmentation results of pancreatic pathological sections.

[0083] In the above technical solution, when scanning the annotated pancreatic pathological sections, the first processing unit fully scans the image-level data and image-level annotations of the annotated pancreatic pathological sections.

[0084] In the above technical solution, the weakly supervised learning framework divides the full scan image into smaller image blocks at equal intervals.

[0085] In the above technical solution, a data transmission channel is established between the first processing unit and the second processing unit.

[0086] The above technical solution realizes the refined processing of the annotated pancreatic pathological sections through the first processing unit and the second processing unit, and can be realized automatically without the need for human participation in the operation of the steps, which not only saves manpower consumption but also has high efficiency. The accuracy of the first processing unit and the second processing unit is improved through training in the first processing unit and the second processing unit, and the errors of annotation and reasoning are effectively reduced. In addition, the pancreatic pathological sections are converted into image segmentation results of the pancreatic pathological sections, which provides convenience for the analysis and processing module, so that the analysis and processing module can perform feature analysis and prediction on the image segmentation results more efficiently and accurately.

[0087] like Figure 5 As shown, in one embodiment provided by the present invention, the first processing unit scans annotated pancreatic pathological sections and uses a combined multi-instance learning method to train and annotate the scanned images, including:

[0088] Fully scan the annotated pancreatic pathology sections to determine image-level data and image-level annotations;

[0089] The combined multi-instance learning method is used to divide the image-level data into multiple image blocks according to the scale factor;

[0090] Sampling all image blocks in the fully scanned annotated pancreatic pathology slice as an image package, and performing judgment and prediction on the image package through a first deep learning module to obtain a first prediction result, and performing judgment and prediction on the image package through a second deep learning module to obtain a second prediction result;

[0091] Eliminate image blocks with different results according to the first prediction result and the second prediction result to obtain a representative image block;

[0092] By training the image block classification model on the representative image block, an image block classification optimization model is obtained;

[0093] The image block classification optimization model is used to perform inference prediction on the image blocks in the image-level data to obtain the prediction results of the hotspot areas of the pancreatic pathological sections;

[0094] Based on the prediction results of hotspot areas in pancreatic pathological sections, the image-level annotations are enhanced to the image block level to obtain the supervision information of the image blocks and obtain the image block-level annotations.

[0095] In the above technical solution, the image blocks are equal in size.

[0096] In the above technical solution, the first deep learning model is a deep learning model based on Max-Max for judgment, and the second deep learning model is a deep learning model based on Max-Min for judgment.

[0097] In the above technical solution, when excluding image blocks with different results based on the first prediction result and the second prediction result, whether the first prediction result and the second prediction result are the same is analyzed based on the same image block, and the image blocks with different first prediction results and the second prediction results are screened out to obtain representative image blocks.

[0098] The above technical solution divides the fully scanned annotated pancreatic pathology slices into smaller image blocks at equal intervals, automatically annotates each block during the supervised information enhancement process, and thus transforms weak supervision into a supervised problem.

[0099] In an embodiment provided by the present invention, after obtaining the image block-level annotation, the first processing unit annotates the image-level data and the image block-level annotation at the pixel level and then outputs them to the second processing unit; wherein, when the image-level data and the image block-level annotation are annotated at the pixel level, the image block-level annotation is directly assigned to each pixel point.

[0100] In the above technical solution, when the image-level data and the image block-level annotations are annotated at the pixel level, a pixel correspondence analysis is performed on the image-level data and the image block-level annotations to obtain a mapping relationship, and the image block-level annotations are directly assigned to each pixel according to the mapping relationship.

[0101] The above technical solution performs pixel-level annotation and then outputs it to the second processing unit, so that the second processing unit can directly perform image segmentation based on pixel points, providing convenience for the second processing unit and improving the efficiency and accuracy of the second processing unit.

[0102] like Figure 6 As shown, in one embodiment provided by the present invention, the second processing unit performs image segmentation model training and reasoning based on pixel-level annotation of image-level data, including:

[0103] Receiving image-level data and pixel-level annotations input by the first processing unit;

[0104] The image segmentation model is trained using image-level data and pixel-level annotations to obtain a trained image segmentation model;

[0105] The image-level data is input into the trained image segmentation model for image segmentation processing to obtain the image segmentation results of pancreatic pathological sections.

[0106] In the above technical solution, when receiving the image-level data and pixel-level annotations input by the first processing unit, the first processing unit receives them in real time, and performs recognition and analysis on the information received in real time to determine the integrity of the received information. When the image-level data and pixel-level annotations input by the first processing unit are complete, the image-level data and pixel-level annotations are used to train the image segmentation model.

[0107] The second processing unit of the above technical solution uses image-level data and pixel-level annotation to train the image segmentation model to improve the accuracy of the image segmentation model, so that the image segmentation model has a smaller error when performing image segmentation processing on image-level data, thereby improving the accuracy of the image segmentation results of pancreatic pathological sections.

[0108] like Figure 7 As shown, in one embodiment provided by the present invention, when the analysis and processing module uses a distributed microprocessor architecture to perform feature analysis and prediction on the image segmentation result, it includes:

[0109] S1. forming a message queue for the pancreatic pathological slice segmentation block in the image segmentation result;

[0110] S2. Performing feature analysis and prediction on the pancreatic pathological slice segmentation block in the working node in turn through the microprocessor according to the message queue to obtain analysis and prediction data of the pancreatic pathological slice segmentation block, including: exploring feature information of the pancreatic pathological slice segmentation block in the image segmentation result from multiple angles to obtain auxiliary diagnosis data of the pancreatic pathological slice segmentation block, and performing prediction based on the auxiliary diagnosis data of the pancreatic pathological slice segmentation block to obtain prediction diagnosis data of the pancreatic pathological slice segmentation block;

[0111] S3. For the analysis and prediction data of the pancreatic pathology section segment block, the auxiliary diagnosis data of the pancreatic pathology section segment block and the prediction and diagnosis data of the pancreatic pathology section segment block are integrated together to obtain the pancreatic pathology auxiliary diagnosis data.

[0112] In the above technical solution, when forming a message queue for the pancreatic pathological slice segments in the image segmentation result, the order of the pancreatic pathological slice segments is determined according to the distribution of the pancreatic pathological slice segments, so as to form the message queue in order.

[0113] In the above technical scheme, when the auxiliary diagnosis data of the pancreatic pathological section segment block is predicted, the prediction is performed in combination with the number of feature information. When the number of feature information is one, historical data information is matched according to the feature information to obtain feature historical associated data information, and then the feature historical associated data information is screened to determine the auxiliary diagnosis data of the pancreatic pathological section segment block. When the number of feature information is two or more, historical data information is matched according to the feature information respectively to obtain feature historical associated data information, and at the same time, feature association combination is performed on the feature information combined with the pathological manifestation to obtain the feature association combination result, and then the feature historical associated data information is combined according to the feature association combination result to obtain the feature historical associated data information, and the auxiliary diagnosis data of the pancreatic pathological section segment block is determined according to the feature historical associated data information.

[0114] In the above technical solution, in the analysis and processing module, the feature analysis and prediction of the image segmentation results are realized through the distributed microprocessor architecture, so that the feature analysis and prediction of the pancreatic pathology slice segmentation blocks in the image segmentation results are carried out in an orderly manner, avoiding omissions or duplications, ensuring the integrity of the pancreatic pathology auxiliary diagnosis data, and exploring the feature information of the pancreatic pathology slice segmentation blocks in the image segmentation results from multiple angles can realize the artificial intelligence analysis of the pancreatic pathology slice segmentation blocks, and efficiently and accurately obtain the auxiliary diagnosis data of the pancreatic pathology slice segmentation blocks from the pancreatic pathology slice segmentation blocks, so that pathologists can make diagnoses based on the auxiliary diagnosis data of the pancreatic pathology slice segmentation blocks, thereby reducing the diagnosis inconsistency caused by the different levels of pathologists, reducing the requirements for pathologists, and improving the accuracy of pathologists' diagnosis. In addition, the distributed microprocessor architecture also has good scalability and fault tolerance, so that the analysis and processing module also has good scalability and fault tolerance.

[0115] In an embodiment provided by the present invention, the number of the working nodes is multiple, and one working node corresponds to one microprocessor. When the microprocessors perform feature analysis and prediction on the pancreatic pathological slice segmentation blocks in the working nodes in sequence according to the message queue, multiple working nodes perform feature analysis and prediction in parallel.

[0116] In the above technical solution, when the feature analysis and prediction of the pancreatic pathological slice segmentation block are performed in the working node through the microprocessor in sequence according to the message queue, the working node is enabled according to the message queue, so that the working node performs feature analysis and prediction of the pancreatic pathological slice segmentation block through the microprocessor.

[0117] The above technical solution enables the feature analysis and prediction of the pancreatic pathological slice segment blocks to be performed in parallel through multiple working nodes, effectively improving the efficiency of the feature analysis and prediction, so that the processing of the pancreatic pathological slice segment blocks can be completed in a shorter time. Moreover, one working node corresponds to one microprocessor. When the feature analysis and prediction of the pancreatic pathological slice segment blocks are performed by the microprocessor, each working node runs as an independent node without affecting each other.

[0118] In one embodiment provided by the present invention, a scheduler is provided in the distributed microprocessor architecture, and the scheduler is used to allocate and manage tasks for working nodes according to a message queue.

[0119] In the above technical solution, when the scheduler is used to allocate and manage tasks for the working nodes according to the message queue, the working nodes are enabled according to the length of the message queue, and the pancreatic pathology section segmentation blocks are allocated and tasks are scheduled for the enabled working nodes through the scheduler.

[0120] The above technical solution uses a scheduler to assign and manage tasks to working nodes according to the message queue, so that the working nodes can perform feature analysis and prediction on the pancreatic pathological slice segmentation blocks in the message queue in an orderly manner, thereby scheduling the working nodes to work according to the progress of the message queue and realizing the overall planning of the working nodes.

[0121] Those skilled in the art should understand that the first and second in the present invention merely refer to different application stages.

[0122] Other embodiments of the present disclosure will be readily apparent to those skilled in the art after considering the specification and practicing the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0123] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An artificial intelligence pancreatic pathology auxiliary diagnosis system, characterized in that: The system includes: a pathology acquisition module, a data annotation module, a supervised learning module and an analysis and processing module; The pathology acquisition module is used to collect and acquire pancreatic pathology slices to obtain pancreatic pathology slices; The data annotation module is used to perform data annotation on the pancreatic pathological sections to obtain annotated pancreatic pathological sections; The supervised learning module is used to scan the annotated pancreatic pathological sections and perform information enhancement and segmentation processing on the scanned images to obtain image segmentation results; The analysis and processing module is used to perform feature analysis and prediction on the image segmentation results using a distributed microprocessing architecture to obtain pancreatic pathology auxiliary diagnosis data.

2. The system according to claim 1, characterized in that The pathology acquisition module includes: an acquisition unit, a preliminary judgment unit and an image preprocessing unit; The acquisition and processing unit is used to acquire pancreatic pathological sections to obtain pancreatic pathological sections; The preliminary judgment unit is used to perform preliminary analysis and judgment on the image quality of the pancreatic pathological slices to obtain preliminary image analysis and judgment results; The image preprocessing unit is used to optimize the pancreatic pathological sections according to the preliminary analysis and judgment results of the images.

3. The system according to claim 1, characterized in that The data annotation module includes: a first annotation unit, a second annotation unit and a screening and review unit; The first labeling unit is used to perform a first labeling on the pancreatic pathological section to obtain a first labeled pancreatic pathological section; The second labeling unit performs a second labeling on the first labeled pancreatic pathology section to obtain a second labeled pancreatic pathology section; The screening and review unit is used to compare and analyze the first annotation and the second annotation in the second annotated pancreatic pathological section, determine whether the first annotation and the second annotation are consistent, obtain analysis results, and screen out pancreatic pathological sections with inconsistent first and second annotations according to the analysis results for further review.

4. The system according to claim 1, characterized in that The supervised learning module uses a weakly supervised learning framework to perform refined processing on the annotated pancreatic pathological sections, and the weakly supervised learning framework includes: a first processing unit and a second processing unit; The first processing unit is used to scan the annotated pancreatic pathological sections, and use a combined multi-instance learning method to train and annotate the scanned images to obtain pixel-level annotations of image-level data; The second processing unit is used to perform image segmentation model training and reasoning based on pixel-level annotation of image-level data to obtain image segmentation results of pancreatic pathological sections.

5. The system according to claim 4, characterized in that The first processing unit scans the annotated pancreatic pathological sections and uses a combined multi-instance learning method to train and annotate the scanned images, including: Fully scan the annotated pancreatic pathology sections to determine image-level data and image-level annotations; The combined multi-instance learning method is used to divide the image-level data into multiple image blocks according to the scale factor; All image blocks are sampled from the fully scanned annotated pancreatic pathological slice as an image package, and a first deep learning module is used to make a judgment and prediction on the image package to obtain a first prediction result, and a second deep learning module is used to make a judgment and prediction on the image package to obtain a second prediction result; Eliminate image blocks with different results according to the first prediction result and the second prediction result to obtain a representative image block; By training the image block classification model on the representative image block, an image block classification optimization model is obtained; The image block classification optimization model is used to perform inference prediction on the image blocks in the image-level data to obtain the prediction results of the hotspot areas of the pancreatic pathological sections; Based on the prediction results of hotspot areas in pancreatic pathological sections, the image-level annotations are enhanced to the image block level to obtain the supervision information of the image blocks and obtain the image block-level annotations.

6. The system according to claim 5, characterized in that After obtaining the image block-level annotation, the first processing unit annotates the image-level data and the image block-level annotation at the pixel level and then outputs them to the second processing unit; wherein, when the image-level data and the image block-level annotation are annotated at the pixel level, the image block-level annotation is directly assigned to each pixel point.

7. The system according to claim 4, characterized in that The second processing unit performs image segmentation model training and reasoning based on pixel-level annotation of image-level data, including: Receiving image-level data and pixel-level annotations input by the first processing unit; The image segmentation model is trained using image-level data and pixel-level annotations to obtain a trained image segmentation model; The image-level data is input into the trained image segmentation model for image segmentation processing to obtain the image segmentation results of pancreatic pathological sections.

8. The system according to claim 1, characterized in that When the analysis and processing module uses a distributed microprocessing architecture to perform feature analysis and prediction on the image segmentation result, it includes: A message queue is formed for the pancreatic pathological slice segmentation block in the image segmentation result; Performing feature analysis and prediction on the pancreatic pathological slice segmentation block in the working node in turn through the microprocessor according to the message queue to obtain analysis and prediction data of the pancreatic pathological slice segmentation block, including: exploring feature information of the pancreatic pathological slice segmentation block in the image segmentation result from multiple angles to obtain auxiliary diagnosis data of the pancreatic pathological slice segmentation block, and performing prediction based on the auxiliary diagnosis data of the pancreatic pathological slice segmentation block to obtain prediction diagnosis data of the pancreatic pathological slice segmentation block; For the analysis and prediction data of the pancreatic pathology section segmentation block, the auxiliary diagnosis data of the pancreatic pathology section segmentation block and the prediction diagnosis data of the pancreatic pathology section segmentation block are integrated together to obtain the pancreatic pathology auxiliary diagnosis data.

9. The system according to claim 8, characterized in that There are multiple working nodes, and one working node corresponds to one microprocessor. When the microprocessors perform feature analysis and prediction on the pancreatic pathological slice segmentation blocks in the working nodes in sequence according to the message queue, multiple working nodes perform feature analysis and prediction in parallel.

10. The system according to claim 9, characterized in that A scheduler is provided in the distributed microprocessor architecture, and the scheduler is used to allocate and manage tasks for working nodes according to the message queue.

Images