Intelligent pathological report automatic generation system based on data analysis

Abstract

The application discloses a wisdom pathological report automatic generation system based on data analysis, relates to the technical field of pathological data management, and comprises a cloud platform, a data acquisition module, a feature processing module, a classification identification module and a report generation module. The data acquisition module constructs a data filtering network to receive pathological section images, the feature processing module performs image processing on the pathological section images to generate pathological image data clusters, judges whether the pathological image data clusters satisfy preset feature identification conditions, extracts pathological section features if yes, and performs secondary image processing if no; the classification identification module inputs all pathological section features into a pathological feature library to identify the pathological information set of the pathological section features, the report generation module generates the pathological report of each patient according to the pathological information set, and stores the pathological report in a diagnosis auxiliary database and intelligently pushes the pathological report.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pathological data management, in particular to a wisdom pathological report automatic generation system based on data analysis. BACKGROUND

[0002] With the continuous development of medical technology, pathological diagnosis plays a crucial role in clinical diagnosis, however, the traditional pathological report generation process relies on the manual operation of pathologists, which has the following problems:

[0003] Low efficiency and high error rate of pathological report generation: pathologists need to spend a lot of time observing pathological sections, so as to write pathological reports, resulting in a long generation cycle of pathological reports, and the manual writing of pathological reports is easily affected by subjective factors, resulting in a decrease in the accuracy of pathological reports.

[0004] High limitation of pathological report application conditions: generally, the doctor who diagnoses the disease of the patient retrieves the corresponding pathological report according to the current disease diagnosis condition of the patient, and a large amount of time of the doctor is used for retrieval work, which indirectly leads to an increase in the workload of disease diagnosis, affecting the diagnosis efficiency.

[0005] Therefore, it is necessary to study a wisdom pathological report automatic generation system based on data analysis to improve the generation efficiency and accuracy of pathological reports, and to push suitable pathological reports to doctors to assist doctors in disease diagnosis. SUMMARY

[0006] In order to solve the above problems, the purpose of the present application is to provide a wisdom pathological report automatic generation system based on data analysis.

[0007] The purpose of the present application can be realized by the following technical scheme: a wisdom pathological report automatic generation system based on data analysis, comprising a cloud platform, the cloud platform being in communication connection with a data acquisition module, a feature processing module, a classification identification module and a report generation module;

[0008] The data acquisition module is used for constructing a data filtering network and receiving pathological section images through the data filtering network;

[0009] The feature processing module is used for image processing of the received pathological section images, and then generating corresponding pathological image data clusters, judging whether the pathological image data clusters meet the preset feature recognition conditions, if yes, extracting pathological section features according to the pathological image data clusters, and if not, performing secondary image processing on the pathological image data clusters;

[0010] The classification and identification module is configured to input all pathological section features into a pathological feature library, and then classify and identify the lesion area, lesion type and lesion degree of each pathological section feature corresponding to a patient, and integrate the same as a pathological information set;

[0011] The report generation module is configured to generate a pathological report of each corresponding patient according to the pathological information set, and store the pathological report in a preset diagnosis assistance database, so that the diagnosis assistance database intelligently pushes the pathological report.

[0012] Further, a data filtering network is constructed, and the process of receiving the pathological section image through the data filtering network comprises:

[0013] The data filtering network is composed of a plurality of data filtering nodes, the plurality of data filtering nodes are numbered and denoted as i, i=1, 2, 3, …, n, n is a natural number greater than 0;

[0014] The data filtering network is configured to receive the pathological section image, each data filtering node in the data filtering network audits the upload IP of the pathological section image, and decides whether to receive the corresponding pathological section image according to the audit result;

[0015] The filtering script corresponding to the data filtering network and the authorized white list are set, the filtering script is used for data filtering of the pathological section image sent by the upload IP which fails in the audit, and the authorized white list is used for recording all authorized upload IPs, and the current pathological section image is received for the authorized upload IP.

[0016] Further, the process of image processing the received pathological section image to generate a corresponding pathological image data cluster comprises:

[0017] The feature processing module is configured to perform image processing on the pathological section image received by the data filtering network, and the image processing comprises image preprocessing, image feature extraction and data clustering;

[0018] The pathological section image is converted into a standard section image through image preprocessing, and the image preprocessing comprises image enhancement, noise filtering and image standardization, wherein the pathological section image is enhanced to improve the contrast and clarity of the pathological section image, the pathological section image is filtered to remove random noise in the image, and the pathological section image is standardized to convert the pathological section image into a preset image size, image brightness and image resolution;

[0019] The standard slice image is divided into a plurality of different slice regions, and the image feature extraction performed on each slice region includes morphological feature extraction, texture feature extraction and color feature extraction, and then the morphological feature data, the texture feature data and the color feature data of each slice region are obtained, which are integrated as the region feature data set of the corresponding slice region;

[0020] According to the region feature data set of the plurality of slice regions corresponding to the standard slice image, data clustering is performed, and then a plurality of region feature data sets within the same data similarity range under a standard slice image are clustered into a pathological image data subcluster of a type corresponding to the current standard slice image. The pathological image data subclusters of the same type in each standard slice image are clustered, and then a plurality of types of pathological image data clusters are generated.

[0021] Further, the process of determining whether the pathological image data cluster meets the preset feature recognition condition includes:

[0022] The feature recognition condition corresponding to the preset pathological image data cluster includes various image parameter thresholds after the pathological slice image is converted into the pathological image data cluster, and also includes an image morphology standard expected to represent the pathological slice image corresponding to the pathological image data cluster;

[0023] Determine whether each pathological image data cluster meets the corresponding feature recognition condition;

[0024] If yes, the corresponding pathological slice feature is extracted according to the current pathological image data cluster;

[0025] If not, the corresponding pathological image data cluster is subjected to secondary image processing.

[0026] Further, the process of extracting the pathological slice feature according to the pathological image data cluster includes:

[0027] A pathological slice feature model is constructed according to historical pathological slice images, and the pathological slice feature model includes a model input layer and a model output layer;

[0028] The model input layer is used to input a plurality of types of pathological image data clusters;

[0029] When the pathological image data cluster is input into the pathological slice feature model, the current pathological image data cluster is matched with the historical pathological slice feature after data processing of the historical pathological slice image of the pathological slice feature model, and the matched historical pathological slice feature is taken as the pathological slice feature of the current corresponding pathological image data cluster;

[0030] The model output layer is configured to output pathological slice features corresponding to each type of pathological image data cluster, and the pathological slice features include cell morphology features, cell structure features, tissue structure features, angiography features, inflammation and lymphatic vessel features, and slice staining features.

[0031] Further, the process of performing secondary image processing on the pathological image data cluster includes:

[0032] The slice region corresponding to the pathological image data cluster that does not meet the feature recognition condition is re-divided, and then the newly generated slice region is reprocessed for image preprocessing, image feature extraction, and data clustering until the newly generated pathological image data cluster meets the feature recognition condition for subsequent extraction of pathological slice features. The pathological image data cluster after secondary image processing is extracted for corresponding pathological slice features.

[0033] Further, the process of inputting all pathological slice features into the pathological feature library and then classifying and identifying the lesion area, lesion type, and lesion degree of each pathological slice feature corresponding to the patient and integrating them as the pathological information set includes:

[0034] The pathological feature library is set, and the pathological feature library stores historical lesion indication data corresponding to historical pathological slice features generated according to historical pathological data, and the historical lesion indication data includes diseases corresponding to a plurality of lesion types that have been identified in historical cases, lesion areas of each lesion type, and lesion degrees of each disease.

[0035] All pathological slice features are input into the pathological feature library, and the pathological feature library compares and matches the historical pathological slice features stored therein with the current pathological slice features. The historical lesion indication data corresponding to the historical pathological slice feature with the highest matching similarity is labeled as the real-time lesion indication of the current pathological slice feature, and then the disease corresponding to each pathological slice feature, the lesion type, the lesion area, and the lesion degree of the disease are classified and identified, and the lesion type, the lesion area, and the lesion degree of the disease are integrated as the corresponding pathological information set.

[0036] Further, the process of generating a pathological report for each corresponding patient according to the pathological information set and storing the pathological report in the preset diagnosis assistance database includes:

[0037] The report generation module performs preliminary data analysis on the pathological information set, and edits the analysis results after data analysis in the form of text into a preset blank file, thereby generating a corresponding preliminary pathological report. The preliminary pathological report records the lesion characteristics, diagnosis conclusion, and treatment suggestions in the form of text.

[0038] The preliminary pathological report is transferred to a professional pathologist, the preliminary pathological report is reviewed by the professional pathologist, the reviewed preliminary pathological report is taken as a final pathological report, the final pathological report is associated with an electronic medical record of a corresponding patient, and then a final pathological report corresponding to each patient is generated, the pathological reports corresponding to all patients are converted into an electronic format, and after being allocated with corresponding search identifiers, the pathological reports are stored in a preset diagnosis assistance database.

[0039] Further, the process of intelligently pushing the pathological report by the diagnosis assistance database comprises:

[0040] The intelligent pushing program is constructed to intelligently push the pathological reports stored in the diagnosis assistance database, the target pushing object of the intelligent pushing is doctors of various departments for diagnosing diseases of patients, the pushing rules corresponding to the intelligent pushing program are defined, the pushing rules are defined based on professional fields of the doctors of various departments, conditions of the patients and data contents of the pathological reports, the intelligent pushing program is executed, the pathological reports in the diagnosis assistance database are pushed to workstations of the doctors for diagnosing diseases of patients, and the doctors are prompted to view the pathological reports;

[0041] The pushing priority of the pathological report by the intelligent pushing program is defined by an algorithm, the pathological reports in the diagnosis assistance database that need to be pushed in priority are automatically identified, whether the pathological report plays an auxiliary role in the current disease diagnosis is evaluated by the doctors after viewing the pathological report, the feedback results of the doctors are collected and returned to the intelligent pushing program as a feedback file, and the pushing rules and the algorithm defined by the current intelligent pushing program are adjusted according to the feedback file.

[0042] Compared with the prior art, the present application has the following advantages:

[0043] 1. The data filtering network is constructed to receive pathological section images meeting the requirements, and the pathological section images received are processed to generate corresponding pathological image data clusters, after judging that the pathological image data clusters meet the preset feature recognition conditions, the corresponding pathological section features are extracted and input into a pathological feature library, the pathological lesion area, the pathological lesion type and the pathological lesion degree of each pathological section feature corresponding to a patient are classified and recognized, and are integrated as a pathological information set, the pathological report corresponding to each patient is generated according to the pathological information set, the generation efficiency and the accuracy of the pathological report are greatly improved, the influence of subjective factors on the generation of the pathological report is effectively avoided, and the error rate of the pathological report is reduced.

[0044] 2. By storing the pathological report into the diagnosis assistance database, constructing the corresponding intelligent pushing program of the diagnosis assistance database, defining the pushing rules of the intelligent pushing program based on the professional fields of doctors in various departments, the conditions of patients and the data content of the pathological report, and defining the pushing priority of the pathological report of the intelligent pushing program through an algorithm, executing the intelligent pushing program, pushing the pathological report to the workstations of a plurality of doctors who diagnose diseases of patients, and prompting the doctors to view, the pathological report is actively pushed, which assists the doctors in diagnosing diseases and helps the doctors in various departments to reduce the data retrieval workload for diagnosing diseases of patients, thereby improving the disease diagnosis efficiency to a certain extent and expanding the application scenarios of the pathological report. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The schematic diagram of the present application. DETAILED DESCRIPTION

[0046] As shown in Figure 1 the intelligent pathological report automatic generation system based on data analysis includes a cloud platform, which is in communication connection with a data acquisition module, a feature processing module, a classification identification module and a report generation module.

[0047] The data acquisition module is used to construct a data filtering network and receive pathological section images through the data filtering network.

[0048] The feature processing module is used to perform image processing on the received pathological section images, and then generate corresponding pathological image data clusters, judge whether the pathological image data clusters meet the preset feature recognition conditions, if yes, extract pathological section features according to the pathological image data clusters, and if not, perform secondary image processing on the pathological image data clusters.

[0049] The classification identification module is used to input all pathological section features into a pathological feature library, and then classify and identify the pathological lesion area, pathological type and pathological degree of each pathological section feature corresponding to the patient, and integrate them as a pathological information set.

[0050] The report generation module is used to generate a pathological report for each corresponding patient according to the pathological information set, and store the pathological report into a preset diagnosis assistance database, and the diagnosis assistance database intelligently pushes the pathological report.

[0051] It should be further explained that in the specific implementation process, the process of constructing the data filtering network and receiving the pathological section images through the data filtering network includes:

[0052] Construct a data filtering network composed of a plurality of data filtering nodes, number the plurality of data filtering nodes, and mark the number as i, then i=1, 2, 3, …, n, where n is a natural number greater than 0;

[0053] The data filtering network is used to receive pathological section images. When the data filtering network starts to work, each data filtering node in the data filtering network audits the upload IP that sends the pathological section images, and decides whether to receive the corresponding pathological section images according to the audit result;

[0054] Set the filtering script corresponding to the data filtering network and the authorized white list;

[0055] The filtering script is used to filter the pathological section images sent by the upload IP that fails the audit;

[0056] The authorized white list is used to record all authorized upload IPs. For the authorized upload IP, the corresponding pathological section image does not need to be filtered, and the current pathological section image is directly received, indicating that the object corresponding to the upload IP that sends the current pathological section image is identified as a legal object;

[0057] When the upload IP that sends the pathological section image is not in the authorized white list, the audit result of the current upload IP is failure, and the pathological section image sent by the current upload IP is removed through data filtering. Otherwise, the audit result is success, and the pathological section image is received for subsequent image processing.

[0058] The object that sends the pathological section image includes a microscope and other medical imaging devices;

[0059] Each data filtering node executes the filtering script corresponding to the data filtering network to perform data filtering, obtains the working parameters of each data filtering node, marks the working parameters of the data filtering node numbered i as Work[i], sets the expected working parameters corresponding to the data filtering node, and marks the expected working parameters as I-Work;

[0060] Mark all data filtering nodes with Work[i]≠I-Work as fault nodes, and repair all fault nodes. Mark all data filtering nodes with Work[i]=I-Work as safe nodes, and do not operate the safe nodes.

[0061] It should be further explained that in the specific implementation process, the received pathological section images are processed to generate corresponding pathological image data clusters, and the process of judging whether the pathological image data clusters meet the preset feature recognition condition includes:

[0062] The pathological section image received by the data filtering network is subjected to image processing by the feature processing module, the image processing including image preprocessing, image feature extraction and data clustering, the image preprocessing, image feature extraction and data clustering being executed in sequence;

[0063] The pathological section image is converted into a standard section image through image preprocessing;

[0064] The image preprocessing includes image enhancement, noise filtering and image standardization;

[0065] The pathological section image is subjected to image enhancement, thereby improving the contrast and definition of the pathological section image, facilitating better identification of the corresponding pathological features in the subsequent process, the pathological section image is subjected to noise filtering to remove random noise in the image, and the pathological section image is subjected to image standardization to convert the pathological section image into a preset image size, image brightness and image resolution;

[0066] The image feature extraction is performed on each standard section image, the standard section image is divided into a plurality of different section regions, and the different region feature data extracted from each section region is integrated as a region feature data set;

[0067] The image feature extraction performed on each section region includes morphological feature extraction, texture feature extraction and color feature extraction, thereby obtaining the region feature data of each section region, the region feature data including morphological feature data, texture feature data and color feature data;

[0068] The morphological feature data, texture feature data and color feature data of each section region are integrated as the corresponding region feature data set;

[0069] The data clustering is performed according to the region feature data sets of the plurality of section regions corresponding to the standard section image, thereby clustering the plurality of region feature data sets within the same data similarity range under one standard section image as a pathological image data subcluster of one type corresponding to the current standard section image;

[0070] The pathological image data subclusters of the same type in each standard section image are clustered, thereby generating a plurality of types of final pathological image data clusters, each type of pathological image data cluster being used to represent the partial pathological related data corresponding to the plurality of patients included in the pathological image data cluster;

[0071] The preset feature recognition condition corresponding to the pathological image data cluster includes various image parameter threshold values after the pathological section image is converted into the pathological image data cluster, the image parameter threshold values include an image resolution threshold value, an image contrast threshold value, and an image saturation threshold value, and further include an image form standard expected by the pathological image data cluster corresponding to the pathological section image;

[0072] It is judged whether each pathological image data cluster meets the corresponding feature recognition condition;

[0073] If yes, it indicates that the current pathological image data cluster has met the subsequent processing requirement, and the corresponding pathological section feature is extracted according to the current pathological image data cluster;

[0074] If no, it indicates that the current pathological image data cluster has not met the subsequent processing requirement, and the secondary image processing is continued on the pathological image data cluster.

[0075] It needs to be further explained that in the specific implementation process, the process of extracting the pathological section feature according to the pathological image data cluster includes:

[0076] A pathological section feature model is constructed according to the historical pathological section image, the pathological section feature model is used for extracting the pathological section feature of the current several types of pathological image data clusters, and the pathological section feature model includes a model input layer and a model output layer;

[0077] The model input layer is used for inputting several types of pathological image data clusters;

[0078] When the pathological image data cluster is input into the pathological section feature model, the current pathological image data cluster is matched with the historical pathological section feature after the historical pathological section image of the pathological section feature model is processed by data, and the matched historical pathological section feature is used as the pathological section feature of the current corresponding pathological image data cluster;

[0079] The model output layer is used for outputting the pathological section feature corresponding to each type of pathological image data cluster, and the pathological section feature includes a cell morphology feature, a cell structure feature, a tissue structure feature, an angiography feature, an inflammation and lymphatic vessel feature, and a section staining feature;

[0080] The cell morphology feature includes the following items:

[0081] Cell size: diameter or area of the cell;

[0082] Cell shape: contour and geometric shape of the cell;

[0083] Nucleus-cytoplasm ratio: area ratio of the cell nucleus and the cytoplasm;

[0084] Nuclear morphology: size, shape, and edge characteristics of nuclei;

[0085] Chromatin structure: distribution, density, and pattern of chromatin within nuclei;

[0086] Cellular arrangement pattern: such as single layer, multiple layers, and aggregation.

[0087] Cellular structural features include the following:

[0088] Cellular border: sharpness of cell borders;

[0089] Cell-to-cell junctions: such as tight junctions and desmosomes;

[0090] Cytoplasmic features: granules, pigments, or other inclusions within cytoplasm.

[0091] Tissue structural features include the following:

[0092] Tissue arrangement: stratification, reticular, or other arrangement of tissue;

[0093] Cell density: number of cells per unit area or volume;

[0094] Interstitial features: composition and structure of tissue interstitium, such as collagen fibers and elastic fibers.

[0095] Angiographic features include the following:

[0096] Vessel density: number of blood vessels per unit area;

[0097] Vessel morphology: branching pattern, lumen size, and other characteristics of blood vessels;

[0098] Inflammatory and lymphatic vessel features include the following:

[0099] Inflammatory cell types: types of cells in areas of inflammation, such as neutrophils and lymphocytes;

[0100] Inflammatory response extent: intensity and extent of inflammatory response;

[0101] Lymphocyte infiltration: distribution and density of cells within lymphatic vessels.

[0102] Section staining features include the following:

[0103] Staining intensity: intensity of coloration of cells or tissue with a particular stain;

[0104] Staining pattern: distribution pattern of stain within cells or tissue.

[0105] The success rate of the pathological slice feature extraction model is obtained in real time, and the success rate is recorded as Sc. The judgment threshold τ is set. When Sc ≥ τ, no operation is performed. When Sc < τ, the historical pathological slice images are used as the training set for the current pathological slice feature model to perform model training.

[0106] The judgment threshold τ is the minimum value corresponding to the success rate of the pathology section feature model in extracting pathology section features. If it is lower than the judgment threshold, it means that the current pathology section feature model needs to be optimized;

[0107] It should be noted that the historical pathology slice features of the historical pathology slice images after data processing are obtained based on the morphological feature data, texture feature data and color feature data generated after the historical pathology slice images are subjected to image feature extraction.

[0108] It should be further explained that, in a specific implementation process, the process of performing secondary image processing on the pathological image data cluster includes:

[0109] Secondary image processing is performed on the pathological image data clusters that do not meet the subsequent processing requirements, and the slice area corresponding to the current pathological image data cluster is re-divided. Then, image preprocessing, image feature extraction and data clustering are re-performed on the newly generated slice area until the regenerated pathological image data cluster meets the subsequent processing requirements for pathological slice feature extraction, and the corresponding pathological slice features are extracted from the pathological image data cluster after the secondary image processing is completed.

[0110] It should be further explained that, in the specific implementation process, all pathological section features are input into the pathological feature library, and then the lesion area, lesion type and lesion severity corresponding to each pathological section feature are classified and identified, and the process of integrating them as a pathological information set includes:

[0111] The classification and recognition module is provided with a pathological feature library, which stores historical lesion indication data corresponding to each of the historical pathological section features generated based on the historical pathological data. The historical lesion indication data includes diseases corresponding to several lesion types that have been identified in the past, the lesion area corresponding to each lesion type, and the lesion degree of each disease;

[0112] The pathological information set is generated according to each corresponding patient, and the pathological report is stored in the preset diagnosis assistance database; the process of intelligently pushing the pathological report by the diagnosis assistance database comprises:

[0113] It should be noted that the live lesion indication is the lesion type, lesion area and lesion degree corresponding to the corresponding pathological section feature.

[0114] It should be further pointed out that in the specific implementation process, the pathological report of each corresponding patient is generated according to the pathological information set, and the pathological report is stored in the preset diagnosis assistance database; the process of intelligently pushing the pathological report by the diagnosis assistance database comprises:

[0115] The report generation module performs preliminary data analysis on the pathological information set, and edits the analysis results after data analysis in the form of text into a preset blank file, thereby generating a corresponding preliminary pathological report; the preliminary pathological report records the lesion characteristics, diagnosis conclusion and treatment suggestion in the form of text;

[0116] The preliminary pathological report is handed over to a professional pathologist for review, and the preliminary pathological report that passes the review is used as the final pathological report; the final pathological report is associated with the electronic medical record of the corresponding patient, thereby generating the final pathological report corresponding to each patient; the pathological report corresponding to all patients is converted into an electronic format, and after being assigned a corresponding search identifier, it is stored in the preset diagnosis assistance database;

[0117] The search identifier is used as the unique search identity certificate of each pathological report;

[0118] A smart push program is constructed, which is used to intelligently push the pathological report stored in the diagnosis assistance database; the target push object of the smart push is each department doctor who diagnoses the disease of the patient;

[0119] The push rule corresponding to the smart push program is defined, which is defined based on the professional field of each department doctor, the condition of the patient and the data content of the pathological report; the smart push program is executed to push the pathological report in the diagnosis assistance database to the workstations of a plurality of doctors who diagnose the disease of the patient, and prompt the doctors to view the pathological report;

[0120] The priority of the push of the pathological report by the intelligent push program is defined by the algorithm, so as to automatically identify the pathological report which needs to be pushed in priority in the diagnosis assistance database;

[0121] The doctor checks the pathological report received by the self workstation, and evaluates whether the pathological report plays an assistance role in the current disease diagnosis, and the feedback result of the doctor is collected as a feedback file and returned to the intelligent push program, the push rule and algorithm of the current intelligent push program are adjusted according to the feedback file, and the push efficiency and effect of the pathological report are optimized, and the workload of the doctors of various departments in the disease diagnosis of the patients is reduced.

[0122] The above examples are only used to illustrate the technical method of the present application, not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. An intelligent pathology report automatic generation system based on data analysis, including a cloud platform, characterized in that: The cloud platform is communicatively connected to a data acquisition module, a feature processing module, a classification and recognition module, and a report generation module; The data acquisition module is used to construct a data filtering network and receive pathological slice images through the data filtering network; The feature processing module is used to perform image processing on the received pathological slice image, thereby generating a corresponding pathological image data cluster, and determining whether the pathological image data cluster meets the preset feature recognition conditions. If so, the pathological slice features are extracted based on the pathological image data cluster; if not, the pathological image data cluster is subjected to secondary image processing; The classification and recognition module is used to input all pathological section features into the pathological feature library, and then classify and identify the lesion area, lesion type and lesion degree of each pathological section feature corresponding to the patient, and integrate them as a pathological information set; The report generation module is used to generate a pathology report for each corresponding patient based on the pathology information set, and store the pathology report in a preset diagnostic auxiliary database, which then intelligently pushes the pathology report. The process of performing image processing on the received pathological slice image to generate a corresponding pathological image data cluster includes: The feature processing module performs image processing on the pathological slice images received by the data filtering network, and the image processing includes image preprocessing, image feature extraction and data clustering; Converting the pathological slice image into a standard slice image through image preprocessing, the image preprocessing includes image enhancement, noise filtering and image standardization, wherein the pathological slice image is enhanced to improve the contrast and clarity of the pathological slice image, the pathological slice image is filtered to remove random noise in the image, and the pathological slice image is standardized to convert the pathological slice image into a preset image size, image brightness and image resolution; The standard slice image is divided into several different slice regions. The image feature extraction for each slice region includes morphological feature extraction, texture feature extraction and color feature extraction, and then the morphological feature data, texture feature data and color feature data of each slice region are obtained and integrated as the regional feature data set of the corresponding slice region. Data clustering is performed based on the regional feature data sets of several slice areas corresponding to the standard slice image, and then several regional feature data sets within the same data similarity range under a standard slice image are clustered into a type of pathological image data sub-cluster corresponding to the current standard slice image. The pathological image data sub-clusters of the same type in each standard slice image are clustered to generate the final several types of pathological image data clusters.

2. The intelligent pathology report automatic generation system based on data analysis according to claim 1 is characterized in that: The process of constructing a data filtering network and receiving pathological slice images through the data filtering network includes: Construct a data filtering network, which consists of a number of data filtering nodes. The data filtering nodes are numbered and denoted as i, where i=1, 2, 3, ..., n, where n is a natural number greater than 0. The data filtering network is used to receive pathological slice images. Each data filtering node in the data filtering network reviews the uploading IP address of the pathological slice image and decides whether to receive the corresponding pathological slice image based on the review result. Set up the filtering script and authorization whitelist corresponding to the data filtering network. The filtering script is used to filter the pathology slice images sent by the upload IP that fails the review. The authorization whitelist is used to record all authorized upload IPs. For the authorized upload IP, the current pathology slice image is received.

3. The intelligent pathology report automatic generation system based on data analysis according to claim 2 is characterized in that: The process of determining whether a pathological image data cluster meets the preset feature recognition conditions includes: Presetting feature recognition conditions corresponding to the pathological image data cluster, the feature recognition conditions including various image parameter thresholds after the pathological slice image is converted into the pathological image data cluster, and also including the image morphology standard corresponding to the pathological image data cluster that represents the expected pathological slice image; Determine whether each pathological image data cluster meets the corresponding feature recognition conditions; If so, the corresponding pathological section features are extracted based on the current pathological image data cluster; If not, secondary image processing is performed on the corresponding pathological image data cluster.

4. The intelligent pathology report automatic generation system based on data analysis according to claim 3 is characterized in that: The process of extracting pathological section features based on pathological image data clusters includes: Constructing a pathology section feature model based on historical pathology section images, the pathology section feature model includes a model input layer and a model output layer; The model input layer is used to input several types of pathological image data clusters; When the pathology image data cluster is input into the pathology slice feature model, the pathology slice feature model matches the current pathology image data cluster with the historical pathology slice features of its own historical pathology slice images after data processing, and uses the successfully matched historical pathology slice features as the pathology slice features of the current corresponding pathology image data cluster; The model output layer is used to output pathological section features corresponding to each type of pathological image data cluster, and the pathological section features include cell morphology features, cell structure features, tissue structure features, angiography features, inflammation and lymphatic vessel features, and section staining features.

5. The intelligent pathology report automatic generation system based on data analysis according to claim 4 is characterized in that: The process of performing secondary image processing on the pathological image data cluster includes: Re-divide the slice areas corresponding to the pathological image data clusters that do not meet the feature recognition conditions, and then re-perform image preprocessing, image feature extraction and data clustering on the newly generated slice areas until the regenerated pathological image data clusters meet the feature recognition conditions for subsequent pathological slice feature extraction, and extract the corresponding pathological slice features from the pathological image data clusters after the secondary image processing.

6. The intelligent pathology report automatic generation system based on data analysis according to claim 5 is characterized in that: The process of inputting all pathological section features into the pathological feature library, classifying and identifying the lesion area, lesion type, and lesion severity corresponding to each pathological section feature, and integrating them into a pathological information set includes: A pathology feature library is set up, wherein the pathology feature library stores historical lesion indication data corresponding to each of the historical pathology section features generated based on the historical pathology data, wherein the historical lesion indication data includes diseases corresponding to several lesion types that have been identified in the past, lesion areas corresponding to each lesion type, and lesion severity of each disease; All pathological section features are input into the pathological feature library, which compares and matches the historical pathological section features stored in itself with the current pathological section features, and marks the historical lesion indication data corresponding to the historical pathological section feature with the highest matching similarity as the actual lesion indication of the current pathological section feature. Then, each pathological section feature is classified and identified to correspond to the disease, lesion type, lesion area and lesion degree of the disease, and the lesion type, lesion area and lesion degree of the disease are integrated as the corresponding pathological information set.

7. The intelligent pathology report automatic generation system based on data analysis according to claim 6 is characterized in that: The process of generating a pathology report for each corresponding patient based on the pathology information set and storing the pathology report in the preset diagnostic auxiliary database includes: The report generation module performs preliminary data analysis on the pathology information set and edits the analysis results in text form into a preset blank file to generate a corresponding preliminary pathology report. The preliminary pathology report records the lesion characteristics, diagnostic conclusions and treatment recommendations in text form; The preliminary pathology report is handed over to a professional pathologist for review. The reviewed preliminary pathology report is used as the final pathology report, and the final pathology report is linked to the electronic medical record of the corresponding patient to generate the final pathology report for each patient. The pathology reports corresponding to all patients are converted into electronic format and stored in the preset diagnostic auxiliary database after being assigned corresponding retrieval identifiers.

8. The intelligent pathology report automatic generation system based on data analysis according to claim 7 is characterized in that: The process of intelligently pushing pathology reports from the diagnostic assistance database includes: Construct a smart push program to push the pathology reports stored in the diagnostic assistance database. The target push objects are doctors from various departments who diagnose patients. Define push rules for the smart push program. The push rules are defined based on the professional fields of doctors in various departments, the patient's condition, and the data content of the pathology report. Execute the smart push program to push the pathology reports in the diagnostic assistance database to the workstations of several doctors who diagnose patients, and prompt the doctors to review the pathology reports. The algorithm defines the push priority of the smart push program for pathology reports, so as to automatically identify the pathology reports that need to be pushed first in the diagnosis assistance database. The doctor will review and evaluate whether the pathology report plays an auxiliary role in the current disease diagnosis. The doctor's feedback results will be collected and returned to the smart push program as a feedback file. The push rules and algorithms defined by the current smart push program will be adjusted according to the feedback file.

Images