Chronic endometritis scoring model development method, scoring model and medium

Abstract

The invention discloses a chronic endometritis scoring model development method, a scoring model and a medium. The method comprises the following steps: collecting chronic endometritis pathological section samples; making a digital pathological image data set based on the chronic endometritis pathological section sample; performing feature extraction based on the digital pathological image data set to obtain pathological features of various chronic endometritis; performing feature fusion on the pathological features to obtain multi-modal features, and constructing a multi-modal feature set; constructing a chronic endometritis scoring model, and training the chronic endometritis scoring model by using the multi-modal feature set to obtain a trained chronic endometritis scoring model; and using the trained chronic endometritis scoring model to assist the diagnosis and treatment of the chronic endometritis. According to the method, the original manual interpretation process is realized through automatic operation, multi-source and multi-dimensional information is effectively integrated, and a more accurate and more objective diagnosis and treatment basis is provided for chronic endometritis.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a chronic endometritis scoring model development method, a scoring model, and a medium. Background Art

[0002] Chronic endometritis is a common gynecological disease, and its clinical diagnosis primarily relies on pathological morphological examination and immunohistochemical staining. Microscopic observation of the microstructural changes in endometrial tissue, combined with the specific expression of immunohistochemical markers, can assist in identifying the characteristics of the inflammatory response. However, this traditional diagnostic model faces significant technical bottlenecks in practical application and urgently needs optimization and improvement.

[0003] The limitations of current technology are mainly reflected in two aspects: First, pathological morphological analysis is highly dependent on manual interpretation. Although pathologists can find characteristic inflammatory manifestations such as plasma cell infiltration through microscopic observation, in actual operation, they often face problems such as vague judgment criteria and strong concealment of subtle lesions, which can easily lead to missed diagnoses and misdiagnoses. Second, there are process defects in immunohistochemistry testing. If multiple antibodies are required for multiple staining verification, the time consumption of manual interpretation of different stained sections will increase significantly, which will not only affect the diagnostic efficiency, but also may lead to a decrease in the timeliness of the inspection due to the lengthy process, and extend the cycle of issuing pathology reports.

[0004] In addition, the subjectivity brought about by manual interpretation of stained sections is unavoidable, resulting in a lack of consistency and repeatability in diagnostic results, affecting the treatment effect. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a chronic endometritis scoring model development method, a scoring model, and a medium.

[0006] A first aspect of the present invention provides a method for developing a chronic endometritis scoring model, comprising the following steps:

[0007] Collecting chronic endometritis pathological section samples; the chronic endometritis pathological section samples at least include HE staining sections and immunohistochemical staining sections;

[0008] Producing a digital pathology image dataset based on the chronic endometritis pathology section sample; the digital pathology image at least includes a HE staining image and an immunohistochemistry staining image;

[0009] Extracting features based on the digital pathological image dataset to obtain pathological features of various chronic endometritis;

[0010] Performing feature fusion on the pathological features of the chronic endometritis to obtain multimodal features reflecting the pathological state of the chronic endometritis, and constructing a multimodal feature set;

[0011] Constructing a chronic endometritis scoring model, and training the chronic endometritis scoring model using the multimodal feature set to obtain a trained chronic endometritis scoring model;

[0012] The trained chronic endometritis scoring model is used to assist in the diagnosis and treatment of chronic endometritis.

[0013] Furthermore, the preparing of the digital pathology image dataset specifically includes: digitally scanning the chronic endometritis pathology slice samples to obtain digital pathology images containing complete slice information, and forming a digital pathology image dataset based on the digital pathology images.

[0014] Furthermore, before the step of preparing the digital pathology image dataset, a step of preprocessing the digital pathology image is also included; the preprocessing step specifically includes:

[0015] performing quality control on the digital pathology image;

[0016] performing color standardization processing on the digital pathology image;

[0017] The digital pathology image is subjected to region segmentation and image block processing.

[0018] Furthermore, the pathological characteristics of chronic endometritis include morphological characteristics of chronic endometritis, and the morphological characteristics are extracted from the HE-stained image by a first deep learning model, specifically comprising the following steps:

[0019] Decomposing the HE-stained image by color channel, identifying the cell structure in the HE-stained image, counting the number and morphology of cell nuclei per unit area, and identifying the density and distribution of glands in the HE-stained image;

[0020] Matching the HE-stained image using a preset specific texture pattern to identify lymphocytes and plasma cells in the HE-stained image;

[0021] quantifying the color intensity and texture complexity of the HE-stained image using a preset gray-level co-occurrence matrix to identify interstitial edema and vascular proliferation in the HE-stained image;

[0022] identifying glandular epithelial changes in the HE-stained image by shape analysis;

[0023] The density and distribution of glands, lymphocytes and plasma cells, interstitial edema and vascular proliferation, and glandular epithelial changes in the HE-stained images are summarized as the morphological characteristics of chronic endometritis.

[0024] Furthermore, the pathological characteristics of chronic endometritis also include immunohistochemical characteristics of chronic endometritis, and the immunohistochemical characteristics are extracted from the immunohistochemical staining image by a second deep learning model, specifically comprising the following steps:

[0025] identifying chronic endometritis marker-positive cells in the immunohistochemical staining image;

[0026] Decomposing the immunohistochemical staining image into multiple image blocks, quantitatively counting the positive cells, and counting the number of positive cells in each image block;

[0027] According to the number of positive cells in each image block, the density and distribution of positive cells in the immunohistochemical staining image are calculated as the immunohistochemical characteristics of chronic endometritis.

[0028] Furthermore, the feature fusion includes at least one of the following steps:

[0029] The pathological features of different chronic endometritis were spliced ​​to obtain the multimodal features of chronic endometritis;

[0030] Different deep learning models with shared network layers are used to jointly learn the pathological features of different chronic endometritis, and the model output is used as the multimodal features of chronic endometritis;

[0031] Deep learning models were trained to process the pathological characteristics of different chronic endometritis types respectively, and the outputs of multiple models were weightedly fused to obtain the multimodal characteristics of chronic endometritis.

[0032] Furthermore, the chronic endometritis scoring model is composed of a regression model; and the training of the chronic endometritis scoring model using the multimodal feature set specifically includes the following steps:

[0033] Dividing the multimodal feature set into a training set and a validation set;

[0034] Performing supervised learning on the chronic endometritis scoring model using the training set and validation set;

[0035] The hyperparameters of the chronic endometritis scoring model are adjusted according to the training results, and the training set and the validation set are used again for training to finally obtain a chronic endometritis scoring model that has completed training.

[0036] Furthermore, after obtaining the trained chronic endometritis scoring model, the method further includes a step of performing a performance evaluation on the trained chronic endometritis scoring model, specifically including the following steps:

[0037] The trained chronic endometritis scoring model was validated using an independent data set; and performance evaluation was performed using the accuracy, precision, recall, F1 score, specificity, and ROC curve of the model output as performance evaluation indicators of the trained chronic endometritis scoring model.

[0038] A second aspect of the present invention provides a chronic endometritis scoring model, which is developed based on the development method described in the first aspect. The chronic endometritis scoring model is operated by the following steps:

[0039] receiving, via a user interface, a digital image of chronic endometritis as input;

[0040] Feature extraction, model inference and score calculation are performed based on the digital image of chronic endometritis to obtain a quantitative score of the digital image of chronic endometritis, and the confidence of the chronic endometritis scoring model is provided for user reference.

[0041] A third aspect of the present invention provides a computer-readable storage medium storing a program, wherein the program is executed by a processor to implement a method for developing a chronic endometritis scoring model as described in the first aspect.

[0042] The present invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the above method.

[0043] The embodiments of the present invention have the following beneficial effects: the present invention provides a chronic endometritis scoring model development method, scoring model and medium, which can automatically learn and accurately identify key morphological features in pathological section images by constructing an artificial intelligence model, reduce the high dependence on manual experience, improve the accuracy and stability of morphological evaluation, and thus reduce the risk of missed diagnosis. The present invention implements the original manual interpretation process through automated operation, greatly shortens the analysis time, significantly improves the diagnostic efficiency and report turnover speed, and eliminates the subjective bias caused by manual interpretation. The present invention can not only perform accurate standardized quantitative analysis of pathological morphological features and multiple immunohistochemical indicators, but also effectively integrate these multi-source and multi-dimensional information based on preset weights or learned intrinsic correlations, and ultimately output a comprehensive, objective and quantitative chronic endometritis score, providing a more accurate and objective basis for clinical diagnosis, grading and efficacy evaluation.

[0044] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0046] Figure 1 This is a basic flow chart of a method for developing a chronic endometritis scoring model according to the present invention.

[0047] Figure 2 It is a schematic diagram of the feature extraction process of morphological features and immunohistochemical features based on digital pathological image datasets in the present invention.

[0048] Figure 3 It is a flowchart of the present invention for training a chronic endometritis scoring model based on a multimodal feature set.

[0049] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0051] In order to solve the technical problems existing in the existing chronic endometritis diagnosis method, the first embodiment of the present invention provides a method for developing a chronic endometritis scoring model. Figure 1 As shown, the following steps are included:

[0052] S1. Collect pathological specimens of chronic endometritis;

[0053] S2. Create a digital pathology image dataset based on chronic endometritis pathology samples;

[0054] S3. Feature extraction based on digital pathology image dataset to obtain pathological features of various chronic endometritis.

[0055] S4. Perform feature fusion on the pathological features of chronic endometritis to obtain multimodal features reflecting the pathological state of chronic endometritis, and construct a multimodal feature set;

[0056] S5. constructing a chronic endometritis scoring model, and training the chronic endometritis scoring model using the multimodal feature set to obtain a trained chronic endometritis scoring model;

[0057] S6. Use the trained chronic endometritis scoring model to assist in the diagnosis and treatment of chronic endometritis.

[0058] The chronic endometritis scoring model development method provided in the embodiments of the present invention can utilize an artificial intelligence model to automate the processing of chronic endometritis, automating the originally time-consuming and labor-intensive manual interpretation process, greatly shortening the analysis time, and significantly improving diagnostic efficiency and report turnover speed; ultimately, a comprehensive, objective and quantitative chronic endometritis score is output, providing a more accurate and objective basis for clinical diagnosis, grading and efficacy evaluation.

[0059] The following describes in detail the implementation process of each step of the present invention:

[0060] S1. Collect pathological samples of chronic endometritis.

[0061] The chronic endometritis pathological section samples collected in the embodiments of the present invention include at least HE staining sections and immunohistochemical staining sections. HE staining sections are cell sections obtained by staining with hematoxylin-eosin, in which the nucleus is blue and the cytoplasm is red. HE staining sections clearly describe the cell structure within the sections and are suitable for extracting the morphological characteristics of chronic endometritis. Immunohistochemical staining sections refer to cell sections that react positively to specific markers (such as CD138, CD38, CD56, MUM-1, etc.), that is, the cells corresponding to the markers in the sections are positive cells. Immunohistochemical staining sections can reflect the spatial distribution, density and other indicators of positive cells, and are suitable for extracting the immunohistochemical characteristics of chronic endometritis. In another part of the embodiments, chronic endometritis pathological section samples also include fluorescent staining sections, microbial staining sections, etc., to meet the different diagnosis and treatment needs of the clinical end.

[0062] S2. Create a digital pathology image dataset based on chronic endometritis pathological section samples.

[0063] In an embodiment of the present invention, a digital pathology image dataset is produced by digitally scanning a pathology section sample of chronic endometritis. Digital scanning can be performed using a full-slice scanner. For example, the pathology section sample can be placed on the scanner's stage, and the section can be precisely moved on the stage using a high-precision mechanical transmission device, so that each area of ​​the section passes through the scanner's objective lens in sequence. At the same time, the scanner's lighting system provides uniform and sufficient light to the section, and the objective lens amplifies the tissue or cell image on the section and focuses it onto an imaging sensor. The imaging sensor is typically a high-resolution CCD or CMOS camera, which converts the optical signal into a digital signal, which is then processed and stored by a computer, ultimately generating a high-resolution digital pathology image of the entire section. For example, scanning a HE-stained section produces an HE-stained image, and scanning an immunohistochemically stained section produces an immunohistochemically stained image.

[0064] After generating digital pathology images, gynecological pathologists need to annotate them according to recognized diagnostic criteria for chronic endometritis (e.g., plasma cell counts within a specific field of view, combined with morphological changes). For example, on hematoxylin and eosinophilic (HE) images, valid endometrial tissue areas are delineated, along with key morphological features such as stromal edema, lymphocytic infiltration, plasma cell aggregates, and glandular structures. On immunohistochemical staining images, the locations of CD138, CD38, and CD56-positive cells are precisely annotated, or positive areas are delineated. Immunohistochemical testing for CD138 and CD38 can help improve the diagnostic yield of chronic endometritis. On CD138 immunohistochemical sections, the presence of five or more plasma cells in hot spots within the endometrial stroma at 400x high-power field of view indicates chronic endometritis. The absence of plasma cells or fewer than five plasma cells per 400x high-power field of view in hot spots indicates non-chronic endometritis. The embodiment of the present invention provides the final diagnosis conclusion and severity classification of each case based on all information as the gold standard for model training. After the annotation is completed, the digital pathology images are combined into a digital pathology image dataset for subsequent model training.

[0065] In some embodiments, before the step of preparing the digital pathology image dataset, a step of preprocessing the digital pathology images is also included; the preprocessing step specifically includes:

[0066] S2-1. Perform quality control on digital pathology images.

[0067] Because pathology slides may contain bubbles or missing cells, and the digital scanning process may also produce image folding and blurring, these defects are considered noise in model training samples. This embodiment of the present invention screens digital pathology images based on preset quality control indicators, eliminating unqualified digital pathology images to prevent them from affecting the quality of subsequent model analysis.

[0068] S2-2. Perform color standardization on digital pathology images.

[0069] Color standardization control of digital pathology images is to ensure that digital pathology images obtained from different slices and different scanners have consistency in color representation, thereby avoiding diagnostic errors caused by color deviation. For example, color space conversion technology (such as converting RGB to HSV or optical density space) can be used to mathematically adjust the hue, saturation, and brightness, and by shifting the mean of the hue distribution to make it close to the standard reference value, and using an algorithm to automatically calculate the color deviation, the balance of each color channel is dynamically corrected to achieve color standardization. In another embodiment, a targeted deep learning model can also be used for color standardization to further improve the comparability between different images.

[0070] S2-3. Perform region segmentation and image block processing on digital pathology images.

[0071] In the embodiments of the present invention, effective endometrial stroma and glandular regions are segmented automatically or semi-automatically, excluding irrelevant areas such as non-tissue background, intracavitary secretions, and large blood vessels. The digital pathology image is segmented into image blocks of a size suitable for input into the deep learning model, which may have overlaps, and the spatial coordinate information of each image block in the original full-slice digitized scan image is recorded.

[0072] S3. Feature extraction was performed based on digital pathological image datasets to obtain the pathological features of various chronic endometritis.

[0073] The embodiment of the present invention mainly performs feature extraction based on a pre-trained deep learning model. Different deep learning models are used for extraction of different types of pathological features. The following is an example of the feature extraction process of the morphological characteristics and immunohistochemical characteristics of chronic endometritis. Figure 2 The pathological features of other types of chronic endometritis were extracted by performing the same or similar process based on their respective deep learning models.

[0074] Morphological features: In the embodiments of the present invention, the morphological features of chronic endometritis are obtained by analyzing preprocessed HE digital images using a first deep learning model. The first deep learning model is preferably a convolutional neural network (CNN), such as ResNet, VGG, the Inception series, or a Vision Transformer. By learning from a large amount of annotated data, the model automatically extracts deep morphological features that reflect the pathological state of chronic endometritis.

[0075] Illustratively, the feature extraction process of morphological features includes the following steps:

[0076] S3-A1. Decompose the HE-stained image by color channel, identify the cell structure in the HE-stained image, count the number and morphology of cell nuclei per unit area, and identify the density and distribution of glands in the HE-stained image;

[0077] S3-A2 uses a preset specific texture pattern to match the HE-stained image and identify lymphocytes and plasma cells in the HE-stained image;

[0078] S3-A3. Quantify the color intensity and texture complexity of HE-stained images using a preset gray-level co-occurrence matrix to identify interstitial edema and vascular proliferation in HE-stained images.

[0079] S3-A4. Identification of glandular epithelial changes in HE-stained images by shape analysis.

[0080] S3-A5. The density and distribution of glands, lymphocytes and plasma cells, interstitial edema and vascular proliferation, and glandular epithelial changes in HE-stained images are summarized as the morphological characteristics of chronic endometritis.

[0081] Immunohistochemical features: The immunohistochemical features of chronic endometritis in the present embodiment are obtained by analyzing preprocessed immunohistochemical digital images using a second deep learning model. This second deep learning model is preferably a structure optimized for cell detection, such as U-Net or Mask R-CNN. It analyzes immunohistochemical staining images for CD138, CD38, and CD56, respectively, to accurately detect and locate cells positive for these markers.

[0082] Illustratively, the feature extraction process of immunohistochemical features includes the following steps:

[0083] S3-B1. Identification of chronic endometritis marker-positive cells in immunohistochemical staining images;

[0084] S3-B2. Decompose the immunohistochemical staining image into multiple image blocks, quantitatively count the positive cells, and count the number of positive cells in each image block;

[0085] S3-B3. Based on the number of positive cells in each image block, the density and distribution of positive cells in immunohistochemical staining images were calculated as the immunohistochemical characteristics of chronic endometritis.

[0086] S4. Perform feature fusion on the pathological features of chronic endometritis to obtain multimodal features reflecting the pathological state of chronic endometritis, and construct a multimodal feature set.

[0087] After completing the extraction of pathological features of chronic endometritis, it is necessary to effectively integrate multiple pathological features including morphological features and immunohistochemical features to form multimodal features that can comprehensively reflect the pathological status of the case.

[0088] The feature fusion strategy includes at least one of the following steps:

[0089] S4-A. Feature splicing of different pathological features of chronic endometritis to obtain multimodal features of chronic endometritis;

[0090] S4-B. Jointly learn the pathological features of different chronic endometritis using different deep learning models with shared network layers. The model output serves as the multimodal signature of chronic endometritis.

[0091] S4-C. Train deep learning models to process the pathological characteristics of different chronic endometritis types separately, perform weighted fusion on the outputs of multiple models, and obtain the multimodal characteristics of chronic endometritis.

[0092] Preferably, a fusion method based on the attention mechanism is adopted to enable the model to adaptively learn the importance of different modal features and different feature dimensions within the same modality for the final diagnosis and scoring, and perform dynamic weighted fusion.

[0093] S5. constructing a chronic endometritis scoring model, and training the chronic endometritis scoring model using the multimodal feature set to obtain a trained chronic endometritis scoring model;

[0094] In embodiments of the present invention, the chronic endometritis scoring model comprises a regression model; the regression model outputs a specific chronic endometritis score for user reference. In other embodiments, a classification model can also be used to output the positive / negative probability of chronic endometritis to achieve scoring. Specific chronic endometritis scoring models can be based on, for example, support vector machines (SVMs), random forests, fully connected neural networks (FCNs), and recurrent neural networks (RNNs) for processing sequence information.

[0095] like Figure 3 As shown in FIG, the chronic endometritis scoring model is trained using a multimodal feature set, which specifically includes the following steps:

[0096] S5-1. Divide the multimodal feature set into a training set and a validation set.

[0097] In some embodiments of the present invention, the dataset is divided into a training set and a validation set in a ratio (e.g., 7:1:2 or 8:1:1). The training set is used to train the model, and the validation set is used to verify the model's generalization ability. In some embodiments, an independent test set can also be divided to perform secondary verification of the model output.

[0098] S5-2. Supervised learning of the chronic endometritis scoring model using training and validation sets.

[0099] Supervised learning is the process of learning a function that can predict or classify new, unseen data by analyzing labeled training data. Its primary goal is to learn patterns from historical data and leverage these patterns to predict or classify future data. In this embodiment of the present invention, supervised learning is performed on the selected model using a training set, with the diagnostic results or scores of image annotations as the training objective.

[0100] S5-3. Adjust the hyperparameters of the chronic endometritis scoring model based on the training results, re-use the training set and validation set for training, and finally obtain a completed chronic endometritis scoring model.

[0101] The embodiment of the present invention monitors the model performance through a verification machine, and prevents overfitting and optimizes the model generalization ability by adjusting hyperparameters (such as learning rate, batch size, number of network layers, number of nodes, etc.), adopting regularization techniques (such as Dropout, L1 / L2 regularization), early stopping and other strategies.

[0102] In some embodiments, after obtaining the trained chronic endometritis scoring model, the method further includes a step of performing a performance evaluation on the trained chronic endometritis scoring model, specifically including the following steps:

[0103] An independent dataset was used to validate the trained chronic endometritis scoring model. The accuracy, precision, recall, F1 score, specificity, and ROC curve of the model output were used as performance evaluation indicators for the trained chronic endometritis scoring model.

[0104] Model performance evaluation can be divided into two aspects: internal validation and external validation. Internal validation refers to the use of a reserved independent test set to conduct a rigorous performance evaluation of the optimal model after training. External validation is to use independent data sets from different medical centers, different time periods or different scanning devices to evaluate the generalization ability and robustness of the model. Performance evaluation indicators may include accuracy, precision, recall, F1 score, specificity, area under the receiver operating characteristic curve (AUC / AUROC), mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R 2 ), etc., the efficiency (e.g., analysis time of a single case), diagnostic accuracy, and consistency (intra-observer and inter-observer differences) of the model-assisted diagnosis were compared and analyzed with those of traditional manual film reading diagnosis, and the scoring effect achieved by the embodiments of the present invention was quantified.

[0105] S6. Use the trained chronic endometritis scoring model to assist in the diagnosis and treatment of chronic endometritis.

[0106] This embodiment of the present invention, based on a validated chronic endometritis scoring model that meets performance standards, converts its output (probability or regression value) into a quantitative score that is easy for clinical staff to understand and use. Specifically, the scoring system can be deployed as a standalone software tool or integrated into a hospital's existing digital pathology information system (DPIS), laboratory information management system (LIS), or electronic medical record system (EMR).

[0107] In another part of the embodiments, trials and feedback collection can also be conducted in actual clinical environments to continuously optimize system performance and user experience, and assist pathologists and clinicians in the rapid and accurate diagnosis, grading, efficacy evaluation, and prognosis prediction of chronic endometritis.

[0108] The artificial intelligence-based chronic endometritis scoring model constructed in the present invention can effectively integrate multi-source pathological information, provide objective, quantitative, and repeatable diagnostic scores, significantly improve diagnostic efficiency and accuracy, reduce subjectivity, and provide solid technical support for the precise diagnosis and treatment of chronic endometritis. It has important clinical application value.

[0109] A second embodiment of the present invention provides a chronic endometritis scoring model, which is developed based on the development method of the first aspect. The chronic endometritis scoring model is operated through the following steps:

[0110] receiving, via a user interface, a digital image of chronic endometritis as input;

[0111] Based on the digital image of chronic endometritis, feature extraction, model inference and score calculation are performed to obtain the quantitative score of the digital image of chronic endometritis, and the confidence of the chronic endometritis scoring model is attached for user reference.

[0112] In general, the embodiments of the present invention address the technical problems in the prior art of manual identification of pathological morphological features, which are highly subjective, difficult, highly dependent on the experience of pathologists, and prone to missed diagnoses. The present invention uses advanced deep learning algorithms to automatically and efficiently learn from complex pathological section images and accurately capture subtle and critical morphological patterns, effectively overcoming the influence of human factors such as experience differences and visual fatigue on the diagnostic results. Through standardized and repeatable feature extraction and analysis, the accuracy of morphological assessment and the consistency between batches and observers are significantly improved, thereby significantly reducing the risk of missed diagnosis and misdiagnosis of chronic endometritis and ensuring the reliability of the diagnostic results. In the face of the technical problems in the prior art of manual interpretation of multiple immunohistochemical staining results such as CD138, CD38, CD56, which are cumbersome, time-consuming and labor-intensive, and seriously affect the turnaround time of pathology reports, the present invention uses an artificial intelligence model to perform fully automatic and high-throughput processing of digitized immunohistochemical staining images. The model can perform rapid and accurate positive cell detection, counting, density calculation and spatial distribution feature analysis on the staining results of multiple markers in parallel or serially. This automated and intelligent analysis process frees pathologists from repetitive and labor-intensive film reading work, greatly compresses the analysis time of a single sample, achieves an order of magnitude improvement in diagnostic efficiency, significantly shortens the pathology report cycle, and strongly supports the need for timely clinical diagnosis and treatment decisions. In response to the technical problems that existing diagnostic methods are inherently subjective, easily lead to significant differences in interpretation between and within observers, and lack stability and comparability of diagnostic results, the present invention uses a deterministic, artificial intelligence-based standardized algorithm to objectively and quantitatively evaluate pathological morphological features and immunohistochemical staining data. The algorithm eliminates the inevitable subjective assumptions and inconsistent standards in human interpretation, and ensures that for the same pathological sample, highly consistent and repeatable quantitative results can be obtained through analysis of this model at different times and under different conditions. This fundamentally improves the objectivity, consistency and repeatability of the diagnosis of chronic endometritis, and lays a technical foundation for establishing a widely recognized and standardized diagnostic process. In response to the technical problems that existing technologies have difficulty in standardizing and accurately quantifying complex pathomorphological information and multiple immunohistochemical indicators, and lack effective means to integrate these heterogeneous, multi-source information to form an overall judgment, the present invention constructs a specific artificial intelligence scoring model. This model can not only accurately and standardizedly analyze the degree and pattern of pathomorphological changes, as well as the expression levels of key immune markers such as CD138, CD38, and CD56, but its core advantage lies in its ability to intelligently and organically integrate these multidimensional information from morphology and multiple molecular markers based on built-in evidence-based medicine logic, preset weights, or the inherent correlation of data mined through machine learning. Ultimately, the model outputs a single, comprehensive, and objectively quantified chronic endometritis score.This score not only accurately reflects the presence and severity of inflammation, but also provides a strong and standardized diagnostic basis for clinicians, supports more refined disease grading, prognosis judgment, treatment plan selection and dynamic monitoring of efficacy, and effectively promotes the precision and individualization of the diagnosis and treatment of chronic endometritis.

[0113] Figure 4 This is a schematic diagram of the structure of the computer-readable storage medium of the third embodiment of the present invention. The computer-readable storage medium of the fourth embodiment of the present invention stores program instructions that can implement the above-mentioned chronic endometritis scoring model development method, wherein the program instructions can be stored in the above-mentioned storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods of each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0114] The contents of the method in the first embodiment of the present invention are all applicable to this computer-readable storage medium embodiment. The functions specifically implemented by this computer-readable storage medium embodiment are the same as those of the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method.

[0115] This embodiment also provides a computer program product. When the computer program product is run on a computer, it enables the computer to execute the above-mentioned related steps to implement the chronic endometritis scoring model development method provided in the above embodiment.

[0116] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0117] Those skilled in the art will appreciate that the modules in the devices in the embodiments of the present invention can be adaptively changed and set in one or more devices different from the embodiments. The modules or units or components in the embodiments of the present invention can be combined into one module or unit or component, and in addition they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including corresponding claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including corresponding claims, abstracts and drawings) can be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0118] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0119] It should be noted that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0120] In addition, each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. In particular, for embodiments such as devices and equipment, since they are basically similar to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The embodiments of the devices and equipment described above are merely schematic, wherein the modules, units, etc. described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed to multiple places, such as nodes in a system network. Specifically, some or all of the modules and units may be selected according to actual needs to achieve the purpose of the above-mentioned embodiment scheme. Those skilled in the art can understand and implement it without paying any creative work.

[0121] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0122] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0123] In addition, the terms "first" and "second" used in the embodiments of the present invention are only used for descriptive purposes and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in this embodiment. Therefore, the features defined by the terms "first" and "second" in the embodiments of the present invention can explicitly or implicitly indicate that the embodiment includes at least one of such features. In the description of the present invention, the word "plurality" means at least two or two or more, such as two, three, four, etc., unless otherwise clearly and specifically defined in the embodiments.

[0124] In the embodiments of the present invention, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or apparatus comprising the element. In addition, components, features, and elements with the same name in different embodiments of the present invention may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context of the specific embodiment.

[0125] Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention, and those of ordinary skill in the art may change, modify, replace, and modify the above embodiments within the scope of the present invention. Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the present invention. This application is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art that are not disclosed in the present invention. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present invention are indicated by the claims below.

Claims

1. A method for developing a scoring model for chronic endometritis, characterized in that: The following steps are involved: Collecting chronic endometritis pathological section samples; the chronic endometritis pathological section samples at least include HE staining sections and immunohistochemical staining sections; Producing a digital pathology image dataset based on the chronic endometritis pathology section sample; the digital pathology image at least includes a HE staining image and an immunohistochemistry staining image; Extracting features based on the digital pathological image dataset to obtain pathological features of various chronic endometritis; Performing feature fusion on the pathological features of the chronic endometritis to obtain multimodal features reflecting the pathological state of the chronic endometritis, and constructing a multimodal feature set; Constructing a chronic endometritis scoring model, and training the chronic endometritis scoring model using the multimodal feature set to obtain a trained chronic endometritis scoring model; The trained chronic endometritis scoring model is used to assist in the diagnosis and treatment of chronic endometritis.

2. The method for developing a chronic endometritis scoring model according to claim 1, wherein: The preparing of the digital pathology image data set specifically includes: digitally scanning the chronic endometritis pathology slice sample to obtain a digital pathology image containing complete slice information, and forming a digital pathology image data set based on the digital pathology image.

3. The method for developing a chronic endometritis scoring model according to claim 1, wherein: Before the step of preparing the digital pathology image data set, the method further includes the step of pre-processing the digital pathology images; The pre-processing step specifically includes: performing quality control on the digital pathology image; performing color standardization processing on the digital pathology image; The digital pathology image is subjected to region segmentation and image block processing.

4. The method for developing a chronic endometritis scoring model according to claim 1, wherein: The pathological characteristics of chronic endometritis include morphological characteristics of chronic endometritis, and the morphological characteristics are extracted from the HE-stained image by a first deep learning model, specifically comprising the following steps: Decomposing the HE-stained image by color channel, identifying the cell structure in the HE-stained image, counting the number and morphology of cell nuclei per unit area, and identifying the density and distribution of glands in the HE-stained image; Matching the HE-stained image using a preset specific texture pattern to identify lymphocytes and plasma cells in the HE-stained image; quantifying the color intensity and texture complexity of the HE-stained image using a preset gray-level co-occurrence matrix to identify interstitial edema and vascular proliferation in the HE-stained image; identifying glandular epithelial changes in the HE-stained image by shape analysis; The density and distribution of glands, lymphocytes and plasma cells, interstitial edema and vascular proliferation, and glandular epithelial changes in the HE-stained images are summarized as the morphological characteristics of chronic endometritis.

5. The method for developing a chronic endometritis scoring model according to claim 1, wherein: The pathological characteristics of chronic endometritis also include immunohistochemical characteristics of chronic endometritis, and the immunohistochemical characteristics are extracted from the immunohistochemical staining image by a second deep learning model, specifically including the following steps: identifying chronic endometritis marker-positive cells in the immunohistochemical staining image; Decomposing the immunohistochemical staining image into multiple image blocks, quantitatively counting the positive cells, and counting the number of positive cells in each image block; According to the number of positive cells in each image block, the density and distribution of positive cells in the immunohistochemical staining image are calculated as the immunohistochemical characteristics of chronic endometritis.

6. The method for developing a chronic endometritis scoring model according to claim 1, wherein: The feature fusion Includes at least one of the following steps: The pathological features of different chronic endometritis were spliced ​​to obtain the multimodal features of chronic endometritis; Different deep learning models with shared network layers are used to jointly learn the pathological features of different chronic endometritis, and the model output is used as the multimodal features of chronic endometritis; Deep learning models were trained to process the pathological characteristics of different chronic endometritis types respectively, and the outputs of multiple models were weightedly fused to obtain the multimodal characteristics of chronic endometritis.

7. The method for developing a chronic endometritis scoring model according to claim 1, wherein: The chronic endometritis scoring model is composed of a regression model; and the multimodal feature set is used to train the chronic endometritis scoring model, specifically comprising the following steps: Dividing the multimodal feature set into a training set and a validation set; Performing supervised learning on the chronic endometritis scoring model using the training set and validation set; The hyperparameters of the chronic endometritis scoring model are adjusted according to the training results, and the training set and the validation set are used again for training to finally obtain a chronic endometritis scoring model that has completed training.

8. The method for developing a chronic endometritis scoring model according to claim 1, wherein: After obtaining the trained chronic endometritis scoring model, the method further includes a step of performing a performance evaluation on the trained chronic endometritis scoring model, specifically including the following steps: The trained chronic endometritis scoring model was validated using an independent data set; and performance evaluation was performed using the accuracy, precision, recall, F1 score, specificity, and ROC curve of the model output as performance evaluation indicators of the trained chronic endometritis scoring model.

9. A chronic endometritis scoring model, developed based on the development method according to any one of claims 1 to 8, characterized in that: The chronic endometritis scoring model is run through the following steps: receiving, via a user interface, a digital image of chronic endometritis as input; Feature extraction, model inference and score calculation are performed based on the digital image of chronic endometritis to obtain a quantitative score of the digital image of chronic endometritis, and the confidence of the chronic endometritis scoring model is provided for user reference.

10. A computer-readable storage medium, characterized in that The storage medium stores a program, and the program is executed by a processor to implement a method for developing a chronic endometritis scoring model according to any one of claims 1 to 8.

Images