Pathological analysis system based on visual inspection

Abstract

The invention belongs to the technical field of medical image processing and medical informatization, and discloses a pathological analysis system based on visual inspection. The system is composed of an image acquisition and intelligent input module, an image feature extraction and index filing module, a parameter combination generation and cancer species adaptation module, a judgment basis and evidence level assignment module, a pathological analysis and dynamic threshold adjustment module, a result generation and multi-platform output module and a result statistics and continuous learning module. According to the method, pathological section images are automatically collected through visual detection equipment, key indexes such as HER2 expression levels, Ki-67 indexes, FISH detection signals, cell sizes and arrangement modes can be efficiently and accurately extracted in combination with an image feature extraction algorithm, and extraction results are automatically filed to corresponding parameter entries; by means of the design, the problems that in the prior art, manual film reading data entry efficiency is low, and manual input is prone to errors are effectively solved, and meanwhile the double-entry mode of photographing automatic item matching and manual additional entry is supported.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing and medical information technology, and specifically relates to a pathology analysis system based on visual detection. Background Art

[0002] The existing pathology analysis process mainly relies on manual reading of films by pathologists and some basic image recognition auxiliary systems.

[0003] Manual film reading, as the current mainstream method of pathological diagnosis, has obvious subjective differences. The reading results are easily affected by human factors such as experience level and fatigue level, resulting in unstable diagnostic accuracy and difficulty in conducting structured analysis of complex parameters.

[0004] Although existing auxiliary diagnosis systems have improved image processing efficiency to a certain extent, they are mostly limited to the extraction of local features from a single image, and are unable to make comprehensive combined judgments on different pathological evidence items. They also lack hierarchical management of dynamic associations between multiple parameters and evidence levels. In addition, existing technologies usually adopt static judgment rules, lack the ability to dynamically adjust thresholds and continuously optimize based on user feedback, and are unable to adapt to the complex needs of parameter changes and sample accumulation in the actual diagnosis process. Especially in the pathological analysis of major diseases such as cancer, a single parameter or single-point interpretation can no longer effectively support clinical precision medication and personalized treatment. Summary of the Invention

[0005] The object of the present invention is to provide a pathology analysis system based on visual detection to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a pathology analysis system based on visual detection, which comprises an image acquisition and intelligent data entry module, an image feature extraction and index archiving module, a parameter combination generation and cancer type adaptation module, a judgment basis and evidence level assignment module, a pathology analysis and dynamic threshold adjustment module, a result generation and multi-platform output module, and a result statistics and continuous learning module; Image acquisition and intelligent data entry module: Pathology slide images are collected through visual inspection equipment, supporting dual-channel simultaneous operation of automatic retrieval and data entry by taking photos and manual data entry, enabling rapid matching of images with pathology parameters, and supporting real-time supplementary recording and editing by users, ensuring efficient and flexible data entry; Image feature extraction and index archiving module: Automatically extract pathological indicators such as HER2, Ki-67, and FISH, support image measurement of parameters such as cell size, nucleolar ratio, and arrangement structure, and archive the extraction results to the parameter library in real time to ensure data integrity, standardization, and accuracy; Parameter combination generation and cancer type adaptation module: Automatically calls the corresponding parameter template based on the type of cancer being tested, generates a combination of primary and auxiliary parameters, and supports personalized parameter adaptation for breast cancer, gastrointestinal cancer, and other solid tumors, ensuring that the parameter combination is scientific and reasonable; Judgment Basis and Evidence Level Assignment Module: This module classifies the generated parameter combinations into evidence levels, labels them as strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence, and automatically assigns confidence levels to each parameter combination to support accurate interpretation. Pathology analysis and dynamic threshold adjustment module: compares detection parameters with the judgment basis set, automatically determines whether it is positive or negative, dynamically adjusts the confidence threshold, and supports real-time optimization based on the false positive rate, missed positive rate, and total sample size to ensure the scientific nature and continuous adaptation of the judgment results; Result generation and multi-platform output module: Generates pathology reports containing test results, parameter combinations, evidence levels, confidence levels, and targeted drug recommendations. It supports multi-format export, batch statistics, and classification display, and supports one-click sharing to WeChat and third-party data platforms. Result statistics and continuous learning module: Continuously records test data and user feedback, automatically compares predictions with actual results, generates error statistics charts, supports user participation in optimization, dynamically adjusts parameter weights and learning cycles based on multi-user data, and continuously improves judgment accuracy.

[0007] Preferably, the image acquisition and intelligent recording module includes: (1) Image acquisition and automatic matching and entry: The pathological sections are photographed and collected by visual inspection equipment. The system automatically extracts image features and quickly retrieves and matches them to corresponding test items based on the built-in pathological parameter index library, including key evidence parameters such as HER2 expression, Ki-67 index, nucleolus size, and cell arrangement. After completing image analysis, the system can automatically and accurately enter the matching test results into the parameter list to ensure efficient and accurate automatic entry, while providing real-time data support for subsequent judgment modules. (2) Manual entry and dual-channel operation are synchronized: To ensure data integrity and correction of abnormal results, the system supports users to supplement or correct parameters through the manual entry unit. Users can select or enter pathological items one by one, such as FISH test results, axillary lymph node metastasis, etc., and support the review of automatically entered results; automatic retrieval and manual entry functions can run synchronously. Users can also manually supplement at any time during the automatic retrieval process of taking photos. The system automatically receives dual-channel entry information in parallel to ensure both entry efficiency and accuracy, while laying the foundation for the system's continuous learning and evidence level judgment.

[0008] Preferably, the image feature extraction and index archiving module includes: (1) Automatic extraction of image features and parameter archiving: Based on the visual inspection equipment, the pathological section images are automatically analyzed. Through the built-in image processing algorithm, key indicators such as HER2 expression level, Ki-67 index, FISH detection signal mean and red-green ratio, cell size, nucleolus size, nucleolus ratio, cell arrangement, nuclear atypia, vascular cancer thrombus and nerve invasion are accurately extracted; the extracted parameters are automatically archived to the system parameter entry library to ensure that the data structure is orderly and the information is clearly classified, providing comprehensive data support for subsequent judgment; (2) Automatic measurement and structured parameter mapping: The cell diameter, nucleolus diameter, signal intensity and cell arrangement structure are automatically measured through image processing algorithms; the system can automatically identify the cell boundaries, nucleolus morphology and arrangement in the image, and quickly complete the quantitative measurement of the corresponding parameters; the measurement results are mapped to the system parameter entry library in real time and automatically archived to the preset entry position to ensure the continuity and accuracy of the parameter extraction process, laying a data foundation for subsequent automatic interpretation and statistical learning of pathological analysis.

[0009] Preferably, the parameter combination generates a cancer type adaptation module: (1) Cancer type identification and parameter template call: After receiving the pathological image and related information, the system can automatically identify the type of cancer corresponding to the detection object, such as breast cancer, gastrointestinal cancer or other solid tumors; based on the identification results, the system automatically calls the parameter template that matches the cancer type to ensure that the called parameter template meets the pathological characteristics and diagnostic requirements of the cancer type; different cancer type parameter templates have built-in settings for main parameters and auxiliary parameters and their combination rules to ensure the pertinence and accuracy of data call; (2) Parameter combination generation and rule matching: Based on the called cancer parameter template, at least one main parameter and one auxiliary parameter are automatically selected, and multiple sets of parameter combinations are generated according to the built-in combination rules; for breast cancer, the parameter combination may include HER2 expression and Ki-67 index, gastrointestinal cancer may involve key indicators such as MSH6 and MSH2, and other tumors call corresponding parameters; the generated parameter combination will be used for subsequent positive or negative judgments to ensure that the interpretation process takes into account both breadth and scientificity.

[0010] Preferably, the judgment basis and evidence level assignment module includes: (1) Parameter combination evidence level classification: The system automatically classifies the generated parameter combinations into four categories based on the clinical diagnostic rules for breast cancer and gastrointestinal cancer: strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence. Strong positive evidence includes HER2 expression level 3+, positive FISH test, large nucleolus, and Ki-67 index higher than 70%. Secondary positive evidence includes HER2 expression level 2+, large nucleolus, Ki-67 index higher than 40%, and the presence of vascular cancer thrombus or nerve invasion. The system quickly classifies all parameter combinations by matching the evidence level library. (2) Probability confidence level assignment and result annotation: For each group of divided parameter combinations, the corresponding probability confidence level is automatically assigned based on the evidence level and historical sample statistical results to measure the reliability of the parameter combination in corresponding positive or negative results. For example, strong positive evidence corresponds to high confidence, and strong negative evidence corresponds to low confidence. At the same time, the system assigns a medium confidence level to positive secondary evidence and negative secondary evidence. All parameter combinations correspond one-to-one with the confidence results and are automatically annotated to the system result library to support subsequent automatic interpretation and clinical reference.

[0011] Preferably, the pathology analysis and dynamic threshold adjustment module includes: (1) Parameter comparison and automatic result judgment: Automatically compare the parameter group of the current test object with the established judgment basis set, and quickly judge whether the test result is positive or negative based on the evidence level and parameter combination confidence level; the comparison process comprehensively considers the main parameters, auxiliary parameters and evidence level to ensure that the judgment process is rigorous and systematic; the preliminary judgment results are recorded in real time and synchronously transmitted to the result statistics and continuous learning module to provide a data basis for subsequent judgment optimization; (2) Dynamic adjustment of confidence threshold: Continuously receive statistical data from the result statistics and the continuous learning module feedback, and dynamically adjust the positive and negative confidence thresholds according to the positive false positive rate and missed positive rate; the adjustment path includes: giving priority to adjusting the threshold when the false positive rate fluctuates greatly; automatically optimizing the threshold update cycle when the total number of samples increases rapidly; through dynamic adjustment, the system can continuously optimize the judgment standard and improve the overall judgment accuracy and adaptability.

[0012] Preferably, the result generation and multi-platform output module includes: (1) Pathology analysis report generation and format support: Based on the automatic judgment results, a complete pathology analysis report containing positive or negative conclusions, parameter combination details, evidence level classification, confidence values ​​and targeted drug recommendation information is generated; the report content is automatically matched according to cancer templates such as breast cancer and gastrointestinal cancer to ensure scientific and rigorous data; the report can be exported to Excel, Word, PDF, and image formats to facilitate long-term retention and multi-scenario application by clinicians, pathologists and patients; (2) Diversified result display and convenient sharing: Supports multiple display modes of pathology analysis reports such as single case export, batch statistical summary, and classification by cancer type to meet the needs of clinical statistics, follow-up management and multi-case comparison; supports one-click sharing function, which can quickly send reports to WeChat, third-party data platforms and other commonly used applications to improve information flow efficiency; the system provides flexible data output paths to facilitate real-time sharing and secure backup of reports between different terminals and platforms.

[0013] Preferably, the result statistics and continuous learning module includes: (1) Continuous recording of test data and error feedback: Continuously record pathological test parameters, predicted results and actual test results input by users, and support automatic comparison of predicted conclusions with actual conditions; for cases with deviations, the system analyzes the prediction errors in real time and generates error statistical charts to facilitate users to intuitively understand the accuracy of the system; users can choose whether to use the error data for subsequent parameter optimization, and the system simultaneously supports batch data import and statistics to ensure that the data accumulation process is continuously effective; (2) Parameter optimization and learning cycle adjustment: Based on continuously recorded data, the parameter weights, evidence priorities, and positive-negative judgment thresholds are dynamically adjusted, and the optimization results are fed back to the pathology analysis module and the dynamic threshold adjustment module in real time. The system supports automatic triggering of parameter optimization and learning cycle adjustment based on the cumulative amount of data input by multiple users, ensuring rapid response when the sample size increases, continuously improving judgment accuracy and adaptability, and ensuring the transparency and rationality of the parameter adjustment process.

[0014] The beneficial effects of the present invention are as follows: 1. The present invention automatically collects pathological section images through visual inspection equipment and combines it with an image feature extraction algorithm to efficiently and accurately extract key indicators such as HER2 expression level, Ki-67 index, FISH detection signal, cell size, and arrangement pattern, and automatically archives the extraction results to corresponding parameter entries. This design effectively solves the problems of low efficiency and error-prone manual data entry in the existing technology. At the same time, it supports a dual entry mode of automatic matching entries by taking photos and manual supplementary entry, which improves the convenience and accuracy of structured management of pathological data and ensures that the parameter basis of subsequent analysis links is more rigorous and reliable.

[0015] 2. The present invention supports the calling of parameter templates across cancer types by designing a dynamic combination mechanism of main parameters and auxiliary parameters, and automatically assigns probability confidence levels to parameter combinations based on the four-level evidence classification rules defined by the customer: strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence. This design breaks through the shortcomings of existing auxiliary diagnosis systems that are limited to single parameters or static combinations. It can provide more objective and quantifiable diagnostic conclusions based on comprehensive judgments of multiple parameters, and support personalized targeted drug recommendations through evidence levels, thereby improving the scientific nature, accuracy and clinical application value of pathological diagnosis.

[0016] 3. The present invention supports users to input real test results and automatically compares errors with the system's predicted results. The system continuously records sample data, feeds back statistical errors, dynamically adjusts the positive and negative judgment thresholds, and optimizes parameter weights and evidence priorities to achieve real-time evolution of diagnostic rules. Compared with existing static diagnostic systems, this system has the ability of continuous learning and self-optimization, and can adapt to parameter fluctuations, sample accumulation and user feedback changes in actual testing environments, effectively reducing the probability of misjudgment and missed judgment, ensuring the timeliness and accuracy of diagnostic results, and providing scientific and reliable technical support for long-term pathological analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the pathology analysis system based on visual detection of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, an embodiment of the present invention provides a pathology analysis system based on visual detection, which is composed of an image acquisition and intelligent input module, an image feature extraction and indicator archiving module, a parameter combination generation and cancer type adaptation module, a judgment basis and evidence level assignment module, a pathology analysis and dynamic threshold adjustment module, a result generation and multi-platform output module, and a result statistics and continuous learning module; Image acquisition and intelligent data entry module: Pathology slide images are collected through visual inspection equipment, supporting dual-channel simultaneous operation of automatic retrieval and data entry by taking photos and manual data entry, enabling rapid matching of images with pathology parameters, and supporting real-time supplementary recording and editing by users, ensuring efficient and flexible data entry; An embodiment of the image acquisition and intelligent data entry module: In actual application, a visual inspection device equipped with a high-resolution camera is used to photograph and capture breast cancer pathological sections. The system immediately initiates an automatic retrieval program after the photo is taken, and through rapid matching of image features, automatically associates the photographed sections with corresponding entries in the parameter entry library, such as HER2 expression, Ki-67 index, and FISH detection. During the photo-taking process, the system simultaneously opens a manual data entry interface, allowing the user to selectively enter parameters that may not be automatically identified, such as vascular cancer thrombus, nerve invasion, and nuclear atypia. This dual-channel synchronous operation mechanism ensures that, based on automatic acquisition, the user can immediately supplement missed parameters, thereby ensuring the integrity and flexibility of data entry.

[0020] Image feature extraction and index archiving module: Automatically extract pathological indicators such as HER2, Ki-67, and FISH, support image measurement of parameters such as cell size, nucleolar ratio, and arrangement structure, and archive the extraction results to the parameter library in real time to ensure data integrity, standardization, and accuracy; An example of the image feature extraction and indicator archiving module: After acquiring a pathology image, the system automatically initiates the image feature extraction algorithm. For breast cancer sections, for example, the system automatically analyzes HER2 expression levels and extracts HER2 positivity by comparing staining depth and regional distribution. Simultaneously, the system measures cell and nucleolar diameters, automatically calculates nucleolar ratios, and extracts cell arrangement patterns (such as solidity, cords, and papillary patterns). For gastrointestinal cancer sections, for example, the system also simultaneously extracts related indicators such as MSH6 and MSH2. Extraction results are archived in real time to a parameter entry library, automatically associating each data item with the sample ID, acquisition time, and detection source, ensuring data integrity and enabling rapid subsequent access.

[0021] Parameter combination generation and cancer type adaptation module: Automatically calls the corresponding parameter template based on the type of cancer being tested, generates a combination of primary and auxiliary parameters, and supports personalized parameter adaptation for breast cancer, gastrointestinal cancer, and other solid tumors, ensuring that the parameter combination is scientific and reasonable; An example of a parameter combination generation and cancer type adaptation module: When the test subject is breast cancer, the system automatically calls the breast cancer parameter template. Based on this template, the system selects at least one primary parameter (such as HER2 expression level) and one auxiliary parameter (such as the Ki-67 index) from core parameters such as HER2, Ki-67, nucleolus size, and cell arrangement to generate a parameter combination. For example, the system automatically generates a combination of "HER2 3 + Ki-67 80%" suitable for a strong positive evidence determination. When the test subject switches to gastrointestinal cancer, the system calls a template containing parameters such as MSH6, PMS2, and MLH1, automatically adjusting the combination rules. Each cancer type template has built-in combination logic to ensure that the parameter combinations are targeted and reasonable.

[0022] Judgment Basis and Evidence Level Assignment Module: This module classifies the generated parameter combinations into evidence levels, labels them as strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence, and automatically assigns confidence levels to each parameter combination to support accurate interpretation. Judgment Basis and Evidence Level Assignment Module Example: During the detection process, the system automatically matches the evidence level rules based on the archived parameters. Taking breast cancer cases as an example, if the HER2 expression level is 3+, the FISH test is positive, the nucleolus is large, and the Ki-67 index is higher than 70%, the system will determine this set of parameters as strong positive evidence and assign a high confidence level (e.g., above 90%). If the parameter combination shows HER2 expression of 2+, the Ki-67 index of 45%, and the presence of vascular tumor thrombus, the system will automatically classify it as secondary positive evidence and assign it a medium confidence level. The system can classify and label all parameter combinations according to the detailed evidence level standards preset in the document, supporting subsequent accurate judgment.

[0023] Pathology analysis and dynamic threshold adjustment module: compares detection parameters with the judgment basis set, automatically determines whether it is positive or negative, dynamically adjusts the confidence threshold, and supports real-time optimization based on the false positive rate, missed positive rate, and total sample size to ensure the scientific nature and continuous adaptation of the judgment results; Pathological analysis and dynamic threshold adjustment module implementation example: The system compares the current parameter group of the test object with the established positive and negative judgment basis set; for example, the current test result is HER22+, Ki-6735%, medium nucleolus, and accompanied by micropapillary structure. The system preliminarily determines it as negative secondary evidence based on the existing judgment rules and outputs a negative result; the system simultaneously retrieves historical data and dynamically adjusts the positive and negative confidence thresholds; if the system continuously finds that the positive false positive rate is increasing, the sensitivity of the positive judgment is reduced by adjusting the threshold; if the total number of samples continues to grow, the system automatically optimizes the threshold update cycle to ensure that real-time adjustment and judgment stability are taken into account.

[0024] Result generation and multi-platform output module: Generates pathology reports containing test results, parameter combinations, evidence levels, confidence levels, and targeted drug recommendations. It supports multi-format export, batch statistics, and classification display, and supports one-click sharing to WeChat and third-party data platforms. Result generation and multi-platform output module implementation example: The system automatically generates a complete pathology analysis report, including the test conclusion (positive or negative), parameter combination details, evidence level classification, current confidence value and a list of recommended targeted drugs (such as trastuzumab); the report can be exported to multiple formats such as Excel, Word, PDF, and images; users can choose to share the report to WeChat, email or a third-party data platform with one click, and support batch export of all test sample reports; the system provides multiple result display methods, supports detailed display of single cases, and also supports batch case statistics and summary by cancer type, to meet the diverse needs of clinical and scientific research.

[0025] Result statistics and continuous learning module: Continuously records test data and user feedback, automatically compares predictions with actual results, generates error statistics charts, supports user participation in optimization, dynamically adjusts parameter weights and learning cycles based on multi-user data, and continuously improves judgment accuracy; Result statistics and continuous learning module implementation example: The system continuously records the parameters, predicted results and actual test conclusions of each pathology test; taking breast cancer cases as an example, after the system gives the predicted results, the user can enter the actual FISH test results, and the system automatically compares and generates error statistics charts; the user can choose whether to include the current error data in the system's continuous learning; if selected, the system automatically adjusts the parameter weights, evidence priority and judgment thresholds; for multiple user inputs, the system automatically starts learning cycle adjustments based on the accumulated sample size to ensure that the parameter optimization rhythm is synchronized with data accumulation, and continuously improve the system's prediction accuracy.

[0026] The image acquisition and intelligent data entry module uses visual inspection equipment to capture and photograph pathological sections. The system automatically extracts image features and, based on a built-in pathological parameter index, quickly retrieves and matches corresponding test items. These include key evidence parameters such as HER2 expression, Ki-67 index, nucleolus size, and cell arrangement. After image analysis, the system automatically and accurately enters the matching test results into the parameter list, ensuring efficient and accurate automatic data entry while providing real-time data support for subsequent judgment modules. To ensure data integrity and correct abnormal results, the system supports user-interpreted or corrected parameters through the manual entry unit. Users can select or enter pathological items one by one, such as FISH test results and axillary lymph node metastasis, and support review of automatically entered results. Automatic retrieval and manual data entry functions can run simultaneously, and users can also manually supplement data at any time during the automatic retrieval process. The system automatically receives dual-channel data in parallel to ensure both data entry efficiency and accuracy, while laying the foundation for continuous learning and evidence level judgment.

[0027] Among them, the image feature extraction and indicator archiving module refers to the automatic analysis of pathological section images based on visual inspection equipment. Through the built-in image processing algorithm, it accurately extracts key indicators including HER2 expression level, Ki-67 index, FISH detection signal mean and red-green ratio, cell size, nucleolus size, nucleolus ratio, cell arrangement, nuclear atypia, vascular cancer thrombus and nerve invasion; the extracted parameters are automatically archived to the system parameter entry library to ensure that the data structure is orderly and the information is clearly classified, providing comprehensive data support for subsequent judgment; the cell diameter, nucleolus diameter, signal intensity and cell arrangement structure are automatically measured through image processing algorithms; the system can automatically identify cell boundaries, nucleolus morphology and arrangement in the image, and quickly complete the quantitative measurement of the corresponding parameters; the measurement results are mapped to the system parameter entry library in real time and automatically archived to the preset entry position, ensuring the continuity and accuracy of the parameter extraction process, laying a data foundation for the subsequent automatic interpretation and statistical learning of pathological analysis.

[0028] Among them, parameter combination generation and cancer type adaptation refers to the ability to automatically identify the type of cancer corresponding to the detection object after receiving pathological images and related information, such as breast cancer, gastrointestinal cancer or other solid tumors; based on the identification results, the system automatically calls the parameter template that matches the cancer type to ensure that the called parameter template meets the pathological characteristics and diagnostic requirements of the cancer type; different cancer parameter templates have built-in settings for main parameters and auxiliary parameters and their combination rules to ensure the targeted and accurate data call; based on the called cancer parameter template, at least one main parameter and one auxiliary parameter are automatically selected, and multiple groups of parameter combinations are generated according to the built-in combination rules; for breast cancer, the parameter combination may include HER2 expression and Ki-67 index, gastrointestinal cancer may involve key indicators such as MSH6 and MSH2, and other tumors call corresponding parameters; the generated parameter combination will be used for subsequent positive or negative judgments to ensure that the interpretation process takes into account both breadth and scientificity.

[0029] The Judgment Basis and Evidence Level Assignment module automatically categorizes generated parameter combinations into four levels of evidence, based on clinical diagnostic rules for breast and gastrointestinal cancers: strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence. Strong positive evidence includes HER2 expression level 3+, positive FISH, large nucleoli, and a Ki-67 index greater than 70%. Secondary positive evidence includes HER2 expression level 2+, large nucleoli, a Ki-67 index greater than 40%, and the presence of vascular tumor thrombus or neural invasion. The system rapidly categorizes all parameter combinations by matching them to the evidence level database. For each categorized parameter combination, the system automatically assigns a probability confidence level based on the evidence level and historical sample statistics to measure the reliability of the parameter combination in generating a positive or negative result. For example, strong positive evidence is assigned a high confidence level, while strong negative evidence is assigned a low confidence level. The system also assigns a medium confidence level to secondary positive evidence and negative secondary negative evidence. All parameter combinations are mapped to confidence levels and automatically annotated in the system's result database to support subsequent automated interpretation and clinical reference.

[0030] Among them, the pathological analysis and dynamic threshold adjustment module refers to automatically comparing the parameter group of the current test object with the established judgment basis set, and quickly judging whether the test result is positive or negative based on the evidence level and the confidence of the parameter combination; the comparison process comprehensively considers the main parameters, auxiliary parameters and evidence level to ensure that the judgment process is rigorous and systematic; the preliminary judgment results are recorded in real time and synchronously transmitted to the result statistics and continuous learning module to provide a data basis for subsequent judgment optimization; the statistical data fed back by the result statistics and continuous learning module are continuously received, and the positive and negative confidence thresholds are dynamically adjusted according to the positive false positive rate and missed judgment rate; the adjustment path includes: giving priority to adjusting the threshold when the false positive rate fluctuates greatly; automatically optimizing the threshold update cycle when the total sample size increases rapidly; through dynamic adjustment, the system can continuously optimize the judgment standard and improve the overall judgment accuracy and adaptability.

[0031] Among them, the result generation and multi-platform output module refers to the generation of a complete pathology analysis report based on automatic judgment results, which includes positive or negative conclusions, parameter combination details, evidence level classification, confidence values ​​and targeted drug recommendation information; the report content is automatically matched according to cancer templates such as breast cancer and gastrointestinal cancer to ensure scientific rigor of the data; the report supports export to Excel, Word, PDF, and image formats, which is convenient for clinicians, pathologists and patients to retain it for a long time and apply it in multiple scenarios; it supports pathology analysis reports in multiple display modes such as single case export, batch statistical summary, and classification by cancer type to meet the needs of clinical statistics, follow-up management and multi-case comparison; it supports one-click sharing function, which can quickly send reports to WeChat, third-party data platforms and other commonly used applications to improve information flow efficiency; the system provides flexible data output paths to facilitate real-time sharing and secure backup of reports between different terminals and platforms.

[0032] Among them, the result statistics and continuous learning module refers to the continuous recording of pathology detection parameters, predicted results and actual test results input by users, and supports automatic comparison of predicted conclusions with actual conditions; for cases with deviations, the system analyzes the prediction errors in real time and generates error statistical charts to facilitate users to intuitively understand the accuracy of the system; users can choose whether to use error data for subsequent parameter optimization, and the system simultaneously supports batch data import and statistics to ensure that the data accumulation process is continuously effective; based on continuously recorded data, the parameter weights, evidence priorities and positive and negative judgment thresholds are dynamically adjusted, and the optimization results are fed back to the pathology analysis module and dynamic threshold adjustment module in real time; the system supports automatic triggering of parameter optimization and learning cycle adjustment based on the cumulative data input by multiple users, ensuring rapid response when the sample size increases, continuously improving judgment accuracy and adaptability, and ensuring the transparency and rationality of the parameter adjustment process.

[0033] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0034] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A pathology analysis system based on visual detection, characterized by: The system consists of an image acquisition and intelligent data entry module, an image feature extraction and index archiving module, a parameter combination generation and cancer type adaptation module, a judgment basis and evidence level assignment module, a pathology analysis and dynamic threshold adjustment module, a result generation and multi-platform output module, and a result statistics and continuous learning module. Image acquisition and intelligent data entry module: Pathology slide images are collected through visual inspection equipment, supporting dual-channel simultaneous operation of automatic retrieval and data entry by taking photos and manual data entry, enabling rapid matching of images with pathology parameters, and supporting real-time supplementary recording and editing by users, ensuring efficient and flexible data entry; Image feature extraction and index archiving module: Automatically extract pathological indicators such as HER2, Ki-67, and FISH, support image measurement of parameters such as cell size, nucleolar ratio, and arrangement structure, and archive the extraction results to the parameter library in real time to ensure data integrity, standardization, and accuracy; Parameter combination generation and cancer type adaptation module: Automatically calls the corresponding parameter template based on the type of cancer being tested, generates a combination of primary and auxiliary parameters, and supports personalized parameter adaptation for breast cancer, gastrointestinal cancer, and other solid tumors, ensuring that the parameter combination is scientific and reasonable; Judgment Basis and Evidence Level Assignment Module: This module classifies the generated parameter combinations into evidence levels, labels them as strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence, and automatically assigns confidence levels to each parameter combination to support accurate interpretation. Pathology analysis and dynamic threshold adjustment module: compares detection parameters with the judgment basis set, automatically determines whether it is positive or negative, dynamically adjusts the confidence threshold, and supports real-time optimization based on the false positive rate, missed positive rate, and total sample size to ensure the scientific nature and continuous adaptation of the judgment results; Result generation and multi-platform output module: Generates pathology reports containing test results, parameter combinations, evidence levels, confidence levels, and targeted drug recommendations. It supports multi-format export, batch statistics, and classification display, and supports one-click sharing to WeChat and third-party data platforms. Result statistics and continuous learning module: Continuously records test data and user feedback, automatically compares predictions with actual results, generates error statistics charts, supports user participation in optimization, dynamically adjusts parameter weights and learning cycles based on multi-user data, and continuously improves judgment accuracy.

2. A pathology analysis system based on visual detection according to claim 1, characterized in that: The image acquisition and intelligent input module includes: (1) Image acquisition and automatic matching and entry: The pathological sections are photographed and collected by visual inspection equipment. The system automatically extracts image features and quickly retrieves and matches them to corresponding test items based on the built-in pathological parameter index library, including key evidence parameters such as HER2 expression, Ki-67 index, nucleolus size, and cell arrangement. After completing image analysis, the system can automatically and accurately enter the matching test results into the parameter list. (2) Manual entry and dual-channel operation are synchronized: To ensure data integrity and correction of abnormal results, the system supports users to supplement or correct parameters through the manual entry unit. Users can select or enter pathology items one by one, and support the review of automatically entered results; automatic retrieval and manual entry functions can run synchronously. Users can also manually supplement at any time during the automatic retrieval process of taking photos. The system automatically receives dual-channel entry information in parallel to ensure both entry efficiency and accuracy.

3. The pathology analysis system based on visual detection according to claim 1, characterized in that: The image feature extraction and index archiving module includes: (1) Automatic image feature extraction and parameter archiving: Based on the visual inspection equipment, the pathological section images are automatically analyzed and the built-in image processing algorithm is used to accurately extract key indicators including HER2 expression level, Ki-67 index, FISH detection signal mean and red-green ratio, cell size, nucleolus size, nucleolus ratio, cell arrangement, nuclear atypia, vascular cancer thrombus and nerve invasion; (2) Automatic measurement and structured parameter mapping: The cell diameter, nucleolus diameter, signal intensity and cell arrangement structure are automatically measured through image processing algorithms; the system can automatically identify the cell boundaries, nucleolus morphology and arrangement in the image, and quickly complete the quantitative measurement of the corresponding parameters; the measurement results are mapped to the system parameter entry library in real time and automatically archived to the preset entry position to ensure the continuity and accuracy of the parameter extraction process.

4. A pathology analysis system based on visual detection according to claim 1, characterized in that: The parameter combination generates a cancer type adaptation module: (1) Cancer type identification and parameter template call: After receiving the pathological image and related information, the system can automatically identify the type of cancer corresponding to the detected object; based on the identification results, the system automatically calls the parameter template that matches the cancer type to ensure that the called parameter template meets the pathological characteristics and diagnostic requirements of the cancer type; different cancer type parameter templates have built-in settings for main parameters and auxiliary parameters and their combination rules to ensure the pertinence and accuracy of data call; (2) Parameter combination generation and rule matching: Based on the called cancer parameter template, at least one main parameter and one auxiliary parameter are automatically selected, and multiple sets of parameter combinations are generated according to the built-in combination rules; for breast cancer, the parameter combination may include HER2 expression and Ki-67 index, gastrointestinal cancer may involve key indicators such as MSH6 and MSH2, and other tumors call corresponding parameters; the generated parameter combination will be used for subsequent positive or negative judgments to ensure that the interpretation process takes into account both breadth and scientificity.

5. The pathology analysis system based on visual detection according to claim 1, characterized in that: The judgment basis and evidence level assignment module includes: (1) Parameter combination evidence level classification: The system automatically classifies the generated parameter combinations into four categories based on the clinical diagnostic rules for breast cancer and gastrointestinal cancer: strong positive evidence, secondary positive evidence, strong negative evidence, and secondary negative evidence. Strong positive evidence includes HER2 expression level 3+, positive FISH test, large nucleolus, and Ki-67 index higher than 70%. Secondary positive evidence includes HER2 expression level 2+, large nucleolus, Ki-67 index higher than 40%, and the presence of vascular cancer thrombus or nerve invasion. The system quickly classifies all parameter combinations by matching the evidence level library. (2) Probability confidence assignment and result annotation: For each group of divided parameter combinations, the corresponding probability confidence is automatically assigned based on the evidence level and historical sample statistical results to measure the reliability of the parameter combination corresponding to the positive or negative result; all parameter combinations correspond one-to-one with the confidence results and are automatically annotated to the system result library to support subsequent automatic interpretation and clinical reference.

6. The pathology analysis system based on visual detection according to claim 1, characterized in that: The pathology analysis and dynamic threshold adjustment module includes: (1) Parameter comparison and automatic result judgment: Automatically compare the parameter group of the current test object with the established judgment basis set, and quickly judge whether the test result is positive or negative based on the evidence level and parameter combination confidence level; the comparison process comprehensively considers the main parameters, auxiliary parameters and evidence level to ensure that the judgment process is rigorous and systematic; the preliminary judgment results are recorded in real time and synchronously transmitted to the result statistics and continuous learning module; (2) Dynamic adjustment of confidence threshold: Continuously receive statistical data from the result statistics and the continuous learning module feedback, and dynamically adjust the positive and negative confidence thresholds according to the positive false positive rate and missed positive rate; the adjustment path includes: giving priority to adjusting the threshold when the false positive rate fluctuates greatly; automatically optimizing the threshold update cycle when the total number of samples increases rapidly; through dynamic adjustment, the system can continuously optimize the judgment standard and improve the overall judgment accuracy and adaptability.

7. The pathology analysis system based on visual detection according to claim 1, characterized in that: The result generation and multi-platform output module includes: (1) Pathology analysis report generation and format support: Based on the automatic judgment results, a complete pathology analysis report containing positive or negative conclusions, parameter combination details, evidence level classification, confidence values ​​and targeted drug recommendation information is generated; the report content is automatically matched according to cancer templates such as breast cancer and gastrointestinal cancer to ensure scientific and rigorous data; the report can be exported to Excel, Word, PDF, and image formats to facilitate long-term retention and multi-scenario application by clinicians, pathologists and patients; (2) Diversified result display and convenient sharing: Supports multiple display modes of pathology analysis reports such as single case export, batch statistical summary, and classification by cancer type to meet the needs of clinical statistics, follow-up management and multi-case comparison; supports one-click sharing function, which can quickly send reports to WeChat, third-party data platforms and other commonly used applications; the system provides flexible data output paths.

8. The pathology analysis system based on visual detection according to claim 1, characterized in that: The result statistics and continuous learning module includes: (1) Continuous recording of test data and error feedback: Continuously record pathological test parameters, predicted results and actual test results input by users, and support automatic comparison of predicted conclusions with actual conditions; for cases with deviations, the system analyzes the prediction errors in real time and generates error statistical charts to facilitate users to intuitively understand the accuracy of the system; users can choose whether to use the error data for subsequent parameter optimization, and the system simultaneously supports batch data import and statistics to ensure that the data accumulation process is continuously effective; (2) Parameter optimization and learning cycle adjustment: Based on continuously recorded data, the parameter weights, evidence priorities, and positive-negative judgment thresholds are dynamically adjusted, and the optimization results are fed back to the pathology analysis module and the dynamic threshold adjustment module in real time. The system supports automatic triggering of parameter optimization and learning cycle adjustment based on the cumulative amount of data input by multiple users, ensuring rapid response when the sample size increases, continuously improving judgment accuracy and adaptability, and ensuring the transparency and rationality of the parameter adjustment process.

Images