Pathology interpretation method, system and readable storage medium based on remote film reading and annotation

Abstract

The present invention discloses a pathology interpretation method, system and readable storage medium based on remote film reading and annotation, wherein the method includes: obtaining an electronic scan of a slide, and digitally displaying the pathology slice information to send to a user end for dual pathology interpretation; inputting the pathology slice information into a trained pathology recognition neural network model to obtain a simulated output result of the model; obtaining the target suspicious area of ​​the pathology slice based on the simulated output result, and synchronously sharing it with the user end; obtaining the two interpretation results returned by the user end and comparing them, and outputting the interpretation results with consistent results as the recognition results of the pathology slice. The present invention improves the accuracy of pathology interpretation by setting up dual pathology interpretation, and traces back to the specific operator. By setting up a neural network model, it can automatically identify areas with high confidence in pathology slices, reduce the repetitive workload of pathologists, and improve the efficiency of pathology recognition.

Description

Technical Field

[0001] The present invention relates to the field of digital recognition technology, and more specifically, to a pathology interpretation method, system and readable storage medium based on remote film reading and annotation. Background Art

[0002] Pathology refers to the process and principle of the occurrence and development of diseases. It refers to the causes and pathogenesis of diseases and the changes and laws in the structure, function and metabolism of cells, tissues and organs that occur during the disease process. Related industries include pathology and pathological sections.

[0003] The current situation in the pathology industry is that it is increasingly difficult to hire pathologists due to the repetitive and mechanized work that can be highly replaced and the boredom. At the same time, in order to complete a large amount of film reading and interpretation work, pathologists are required to sit in front of microscopes for long periods of time, which causes great personal harm to pathologists and invisibly increases investment costs. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide a pathology interpretation method, system and readable storage medium based on remote film reading and annotation, which can realize digital remote reading of pathology sections. By setting up dual pathology interpretation, the interpretation accuracy can be effectively improved, and a neural network model is set for automatic recognition, which can reduce the repetitive work of pathologists, thereby improving the interpretation efficiency of pathology sections.

[0005] The first aspect of the present invention provides a pathology interpretation method based on remote film reading and annotation, comprising the following steps:

[0006] Obtain electronic scans of slides and digitally display pathological section information to send to the user end for dual pathological interpretation;

[0007] Inputting the pathological section information into a trained pathology recognition neural network model to obtain a simulated output result of the model;

[0008] Acquire the target suspicious area of ​​the pathological slice based on the simulation output result, and synchronously share it with the user terminal;

[0009] The two interpretation results returned by the user end are obtained and compared, and the interpretation results that are consistent with each other are output as the recognition results of the pathological sections.

[0010] In this solution, the electronic scan of the slide is obtained, and the pathological section information is digitally displayed to be sent to the user end for dual pathological interpretation, specifically:

[0011] Scanning the glass slide with a preset scanner to obtain the electronic scanned copy;

[0012] Matching the electronic scanned document based on a pre-stored slice database to obtain the pathological slice information, and digitally displaying the pathological slice information;

[0013] The same pathological slice information is sent to the user end twice for pathological judgment, so as to obtain a test judgment result and an audit judgment result returned by the user end.

[0014] In this solution, the pathology recognition neural network model training method is:

[0015] Obtain pathological slide information and slide recognition results of historical slide scans;

[0016] Preprocessing the pathological section information and the section recognition results of the historical slide scans to obtain a training sample set;

[0017] Inputting the training sample set into the initialized pathology recognition neural network model for training;

[0018] Get the accuracy of the output results;

[0019] If the accuracy rate is greater than a preset accuracy rate threshold, the training is stopped to obtain the pathology recognition neural network model.

[0020] In this solution, the target suspicious area of ​​the pathological slice is obtained based on the simulation output result, and is synchronously shared with the user terminal, specifically:

[0021] Extracting the simulation output results, calling the simulation output database to identify the pathological section analysis results;

[0022] Obtaining a target suspicious area of ​​the pathological section based on the analysis result of the pathological section;

[0023] The pathological slice attribute information is obtained for identity recognition, and the target suspicious area is synchronized with the pathological slice information of the corresponding identity for updating, so as to provide a reference for the user to perform pathological interpretation.

[0024] In this solution, the two interpretation results returned by the user end are obtained and compared, and the interpretation results with consistent results are output as the recognition results of the pathological sections, specifically:

[0025] Obtaining the test interpretation result and the review interpretation result marked by the user;

[0026] Calculate the offset of the two reading results, where:

[0027] If the offset is less than or equal to a preset threshold, the comparison result is determined to be consistent;

[0028] If the offset is greater than the preset threshold, the comparison result is determined to be inconsistent;

[0029] The audit judgment result that is compared and found to be consistent is output as the recognition result of the pathological section.

[0030] This solution also includes counting the workload of the user end, specifically:

[0031] Defining the output of the recognition result of the pathological section once as a workload;

[0032] Counting and quantifying the total workload of the user terminal based on a preset period;

[0033] A corresponding remuneration mechanism is matched based on the total workload value to output a total remuneration value.

[0034] The second aspect of the present invention further provides a pathology interpretation system based on remote film reading and annotation, comprising a memory and a processor, wherein the memory comprises a pathology interpretation method program based on remote film reading and annotation, and when the pathology interpretation method program based on remote film reading and annotation is executed by the processor, the following steps are implemented:

[0035] Obtain electronic scans of slides and digitally display pathological section information to send to the user end for dual pathological interpretation;

[0036] Inputting the pathological section information into a trained pathology recognition neural network model to obtain a simulated output result of the model;

[0037] Acquire the target suspicious area of ​​the pathological slice based on the simulation output result, and synchronously share it with the user terminal;

[0038] The two interpretation results returned by the user end are obtained and compared, and the interpretation results that are consistent with each other are output as the recognition results of the pathological sections.

[0039] In this solution, the electronic scan of the slide is obtained, and the pathological section information is digitally displayed to be sent to the user end for dual pathological interpretation, specifically:

[0040] Scanning the glass slide with a preset scanner to obtain the electronic scanned copy;

[0041] Matching the electronic scanned document based on a pre-stored slice database to obtain the pathological slice information, and digitally displaying the pathological slice information;

[0042] The same pathological slice information is sent to the user end twice for pathological judgment, so as to obtain a test judgment result and an audit judgment result returned by the user end.

[0043] In this solution, the pathology recognition neural network model training method is:

[0044] Obtain pathological slide information and slide recognition results of historical slide scans;

[0045] Preprocessing the pathological section information and the section recognition results of the historical slide scans to obtain a training sample set;

[0046] Inputting the training sample set into the initialized pathology recognition neural network model for training;

[0047] Get the accuracy of the output results;

[0048] If the accuracy rate is greater than a preset accuracy rate threshold, the training is stopped to obtain the pathology recognition neural network model.

[0049] In this solution, the target suspicious area of ​​the pathological slice is obtained based on the simulation output result, and is synchronously shared with the user terminal, specifically:

[0050] Extracting the simulation output results, calling the simulation output database to identify the pathological section analysis results;

[0051] Obtaining a target suspicious area of ​​the pathological section based on the analysis result of the pathological section;

[0052] The pathological slice attribute information is obtained for identity recognition, and the target suspicious area is synchronized with the pathological slice information of the corresponding identity for updating, so as to provide a reference for the user to perform pathological interpretation.

[0053] In this solution, the two interpretation results returned by the user end are obtained and compared, and the interpretation results with consistent results are output as the recognition results of the pathological sections, specifically:

[0054] Obtaining the test interpretation result and the review interpretation result marked by the user;

[0055] Calculate the offset of the two reading results, where:

[0056] If the offset is less than or equal to a preset threshold, the comparison result is determined to be consistent;

[0057] If the offset is greater than the preset threshold, the comparison result is determined to be inconsistent;

[0058] The audit judgment result that is compared and found to be consistent is output as the recognition result of the pathological section.

[0059] This solution also includes counting the workload of the user end, specifically:

[0060] Defining the output of the recognition result of the pathological section once as a workload;

[0061] Counting and quantifying the total workload of the user terminal based on a preset period;

[0062] A corresponding remuneration mechanism is matched based on the total workload value to output a total remuneration value.

[0063] The third aspect of the present invention provides a computer-readable storage medium, which includes a machine-based pathology interpretation method program based on remote film reading and annotation. When the pathology interpretation method program based on remote film reading and annotation is executed by a processor, the steps of a pathology interpretation method based on remote film reading and annotation as described in any one of the above items are implemented.

[0064] The present invention discloses a pathology interpretation method, system and readable storage medium based on remote film reading and annotation. Double pathology interpretation is set to improve the accuracy of pathology interpretation and trace back to the specific operator. A neural network model is set to automatically identify areas with high confidence in pathology sections, thereby reducing the repetitive workload of pathologists and improving the efficiency of pathology identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A flow chart of a pathology interpretation method based on remote film reading and annotation according to the present invention is shown;

[0066] Figure 2 A block diagram of a pathology interpretation system based on remote film reading and annotation of the present invention is shown. DETAILED DESCRIPTION

[0067] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0068] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0069] Figure 1 A flow chart of a pathology interpretation method based on remote film reading and annotation of the present application is shown.

[0070] like Figure 1 As shown, the present application discloses a pathology interpretation method based on remote film reading and annotation, comprising the following steps:

[0071] S102, obtaining an electronic scan of the slide and digitally displaying the pathological slice information to send to the user end for dual pathological interpretation;

[0072] S104, inputting the pathological section information into a trained pathology recognition neural network model to obtain a simulation output result of the model;

[0073] S106, acquiring a target suspicious area of ​​the pathological slice based on the simulation output result, and synchronously sharing it with the user terminal;

[0074] S108, obtaining and comparing the two interpretation results returned by the user terminal, and outputting the interpretation results that are consistent as the recognition results of the pathological sections.

[0075] It should be noted that, first, the electronic scan of the slide is obtained, and the corresponding pathological section information is digitally displayed and sent to the user end for interpretation. A dual interpretation mechanism is set up, which is a test interpretation and an audit interpretation. The effectiveness of the work of the pathologist is evaluated through the two interpretation results. The pathological section information is input into the trained pathological recognition neural network model to obtain the simulated output result of the model, and then the target suspicious area of ​​the pathological section is obtained and synchronously shared with the user end, so as to speed up the interpretation time of the pathologist. The interpretation time can be shortened from "2-3" min to "about 30 s" on average. The two interpretation results are compared, and the interpretation result with consistent results is output as the recognition result of the pathological section.

[0076] According to an embodiment of the present invention, the electronic scan of the slide is obtained, and the pathological section information is digitally displayed to be sent to the user end for dual pathological interpretation, specifically:

[0077] Scanning the glass slide with a preset scanner to obtain the electronic scanned copy;

[0078] Matching the electronic scanned document based on a pre-stored slice database to obtain the pathological slice information, and digitally displaying the pathological slice information;

[0079] The same pathological slice information is sent to the user end twice for pathological judgment, so as to obtain a test judgment result and an audit judgment result returned by the user end.

[0080] It should be noted that the electronic scan is obtained by scanning the glass slide using a TDI-CDD linear scanner, the electronic scan is matched based on a pre-stored slice database to obtain the pathological slice information, and the pathological slice information is digitally displayed. The slice database includes a rectangular wireframe, a circular wireframe, a triangular wireframe and a breakpoint database.

[0081] According to an embodiment of the present invention, the target suspicious area of ​​the pathological slice is obtained based on the simulation output result, and is synchronously shared with the user terminal, specifically:

[0082] Extracting the simulation output results, calling the simulation output database to identify the pathological section analysis results;

[0083] Obtaining a target suspicious area of ​​the pathological section based on the analysis result of the pathological section;

[0084] The pathological slice attribute information is obtained for identity recognition, and the target suspicious area is synchronized with the pathological slice information of the corresponding identity for updating, so as to provide a reference for the user to perform pathological interpretation.

[0085] It should be noted that the attribute information of the pathological section is first obtained and encoded one-to-one, so that each pathological section can be tracked to an individual, ensuring the complete interpretation chain of the pathologist and the pathological section, calling the simulated output database to identify the pathological section analysis results, and obtaining the target suspicious area of ​​the pathological section based on the pathological section analysis results, and then synchronizing the target suspicious area to the pathological section information of the corresponding identity for updating, so that the pathologist on the user side can find the problem at the first time and conduct focused analysis and interpretation.

[0086] According to an embodiment of the present invention, the obtaining and comparing of the two interpretation results returned by the user terminal, and outputting the interpretation results that are consistent with each other as the recognition results of the pathological sections, specifically includes:

[0087] Obtaining the test interpretation result and the review interpretation result marked by the user;

[0088] Calculate the offset of the two reading results, where:

[0089] If the offset is less than or equal to a preset threshold, the comparison result is determined to be consistent;

[0090] If the offset is greater than the preset threshold, the comparison result is determined to be inconsistent;

[0091] The audit judgment result that is compared and found to be consistent is output as the recognition result of the pathological section.

[0092] It should be noted that the interpretation results correspond to different ratios of disease causes and pathogenesis, so the offset θ of the interpretation results needs to be calculated twice, and the calculation formula is as follows:

[0093]

[0094] Among them, α and β are dynamic parameters, which are specified manually.1 The test result, R 2 For the review and interpretation result, the preset threshold is taken as 5%, wherein, if the offset "θ≤0.05", the result comparison is determined to be consistent, and the review and judgment result is output as the recognition result of the pathological section; if the offset "θ>0.05", the result comparison is determined to be inconsistent and is not output.

[0095] According to an embodiment of the present invention, the pathology recognition neural network model training method is:

[0096] Obtain pathological slide information and slide recognition results of historical slide scans;

[0097] Preprocessing the pathological section information and the section recognition results of the historical slide scans to obtain a training sample set;

[0098] Inputting the training sample set into the initialized pathology recognition neural network model for training;

[0099] Get the accuracy of the output results;

[0100] If the accuracy rate is greater than a preset accuracy rate threshold, the training is stopped to obtain the pathology recognition neural network model.

[0101] It should be noted that the pathology recognition neural network model requires a large amount of historical data for training. The larger the amount of data, the more accurate the result. The pathology recognition neural network model in this application can be trained by using the pathology slice information and slice recognition results of the historical slide scans as input. Of course, when training the neural network model, it is necessary not only to train with the pathology slice information and slice recognition results of the historical slide scans, but also to train in combination with the determined pathology slice disease causes and pathogenesis. By comparing a large amount of test data with real data, the results obtained will be more accurate, thereby making the output results of the pathology recognition neural network more accurate. Preferably, the accuracy threshold is generally set to 90%.

[0102] According to an embodiment of the present invention, it also includes counting the workload of the user terminal, specifically:

[0103] Defining the output of the recognition result of the pathological section once as a workload;

[0104] Counting and quantifying the total workload of the user terminal based on a preset period;

[0105] A corresponding remuneration mechanism is matched based on the total workload value to output a total remuneration value.

[0106] It should be noted that the workload statistics of the pathologist can also be obtained by counting the output results, counting the output results within a preset period (for example, 30 days), and quantifying them as the total workload of the user end to match the corresponding remuneration mechanism to output the total remuneration value. For example, the total workload value is "100" and the corresponding total remuneration value is "1500"; it is worth mentioning that only the output review and judgment result as the identification result of the pathological section can become a workload, thereby ensuring the work quality of the pathologist.

[0107] Furthermore, when allocating tasks, it also includes matching users according to task label information and completing task scheduling to improve the overall task processing efficiency and accuracy. Simply put, it is to allow more suitable users to complete the corresponding tasks. Among them, for the scheduling of processing time: the pathological slices are marked with time labels to obtain the urgency of the processing, and the priority tasks are scheduled accordingly. For example, users with faster average task response and shorter processing time can receive more tasks, and more urgent tasks will be distributed to such users first. For the scheduling of task matching, due to the different abilities and advantages of different pathologists, it can be judged based on past task processing results, and more suitable tasks can be recommended to doctors. For example, according to past results, doctor A has an accuracy rate of more than 80% in the interpretation of HSIL slices. Then, when AI processing considers suspected HSIL, the reading task of the slice will be distributed to doctor A for processing first.

[0108] It is worth mentioning that the pathological interpretation method based on remote film reading and annotation proposed in this application also includes grading the marked target suspicious area, specifically:

[0109] Acquire the target suspicious area of ​​the pathological section;

[0110] Counting the number of slice data matched by the target suspicious area;

[0111] The proportion of the number of slice data values ​​can be calculated based on the target area, wherein:

[0112] If the proportion exceeds the level threshold, the level of the target suspicious area is classified as the first level;

[0113] If the proportion does not exceed the level threshold, the level of the target suspicious area is classified as the second level.

[0114] It should be noted that the degree of change of the diseased cells in the target suspicious area of ​​the pathological section is different, so it is necessary to count the matching slice data in the target suspicious area to judge the degree of the lesion accordingly. The level threshold is taken as 60%. If the proportion exceeds 60%, it indicates that the degree of diseased cells in the target suspicious area has exceeded 60%. The level of the target suspicious area is divided into the first level. When the user-side pathologist interprets it, it can be interpreted according to different level divisions.

[0115] Figure 2 A block diagram of a pathology interpretation system based on remote film reading and annotation of the present invention is shown.

[0116] like Figure 2 As shown, the present invention discloses a pathology interpretation system based on remote film reading and annotation, including a memory and a processor, wherein the memory includes a pathology interpretation method program based on remote film reading and annotation, and when the pathology interpretation method program based on remote film reading and annotation is executed by the processor, the following steps are implemented:

[0117] Obtain electronic scans of slides and digitally display pathological section information to send to the user end for dual pathological interpretation;

[0118] Inputting the pathological section information into a trained pathology recognition neural network model to obtain a simulated output result of the model;

[0119] Acquire the target suspicious area of ​​the pathological slice based on the simulation output result, and synchronously share it with the user terminal;

[0120] The two interpretation results returned by the user end are obtained and compared, and the interpretation result that passes the preset review pass rate is output as the recognition result of the pathological section.

[0121] It should be noted that, first, the electronic scan of the slide is obtained, and the corresponding pathological section information is digitally displayed and sent to the user end for interpretation. A dual interpretation mechanism is set up, which is a test interpretation and an audit interpretation. The effectiveness of the work of the pathologist is evaluated through the two interpretation results. The pathological section information is input into the trained pathological recognition neural network model to obtain the simulated output result of the model, and then the target suspicious area of ​​the pathological section is obtained and synchronously shared with the user end, so as to speed up the interpretation time of the pathologist. The interpretation time can be shortened from "2-3" min to "about 30 s" on average. The two interpretation results are compared, and the interpretation result with consistent results is output as the recognition result of the pathological section.

[0122] According to an embodiment of the present invention, the electronic scan of the slide is obtained, and the pathological section information is digitally displayed to be sent to the user end for dual pathological interpretation, specifically:

[0123] Scanning the glass slide with a preset scanner to obtain the electronic scanned copy;

[0124] Matching the electronic scanned document based on a pre-stored slice database to obtain the pathological slice information, and digitally displaying the pathological slice information;

[0125] The same pathological slice information is sent to the user end twice for pathological judgment, so as to obtain a test judgment result and an audit judgment result returned by the user end.

[0126] It should be noted that the electronic scan is obtained by scanning the glass slide using a TDI-CDD linear scanner, the electronic scan is matched based on a pre-stored slice database to obtain the pathological slice information, and the pathological slice information is digitally displayed. The slice database includes a rectangular wireframe, a circular wireframe, a triangular wireframe and a breakpoint database.

[0127] According to an embodiment of the present invention, the target suspicious area of ​​the pathological slice is obtained based on the simulation output result, and is synchronously shared with the user terminal, specifically:

[0128] Extracting the simulation output results, calling the simulation output database to identify the pathological section analysis results;

[0129] Obtaining a target suspicious area of ​​the pathological section based on the analysis result of the pathological section;

[0130] The pathological slice attribute information is obtained for identity recognition, and the target suspicious area is synchronized with the pathological slice information of the corresponding identity for updating, so as to provide a reference for the user to perform pathological interpretation.

[0131] It should be noted that the attribute information of the pathological section is first obtained and encoded one-to-one, so that each pathological section can be tracked to an individual, ensuring the complete interpretation chain of the pathologist and the pathological section, calling the simulated output database to identify the pathological section analysis results, and obtaining the target suspicious area of ​​the pathological section based on the pathological section analysis results, and then synchronizing the target suspicious area to the pathological section information of the corresponding identity for updating, so that the pathologist on the user side can find the problem at the first time and conduct focused analysis and interpretation.

[0132] According to an embodiment of the present invention, the obtaining and comparing of the two interpretation results returned by the user terminal, and outputting the interpretation results that are consistent with each other as the recognition results of the pathological sections, specifically includes:

[0133] Obtaining the test interpretation result and the review interpretation result marked by the user;

[0134] Calculate the offset of the two reading results, where:

[0135] If the offset is less than or equal to a preset threshold, the comparison result is determined to be consistent;

[0136] If the offset is greater than the preset threshold, the comparison result is determined to be inconsistent;

[0137] The audit judgment result that is compared and found to be consistent is output as the recognition result of the pathological section.

[0138] It should be noted that the interpretation results correspond to different ratios of disease causes and pathogenesis, so the offset θ of the interpretation results needs to be calculated twice, and the calculation formula is as follows:

[0139]

[0140] Among them, α and β are dynamic parameters, which are specified manually. 1 The test result, R 2 For the review and interpretation result, the preset threshold is taken as 5%, wherein, if the offset "θ≤0.05", the result comparison is determined to be consistent, and the review and judgment result is output as the recognition result of the pathological section; if the offset "θ>0.05", the result comparison is determined to be inconsistent and is not output.

[0141] According to an embodiment of the present invention, the pathology recognition neural network model training method is:

[0142] Obtain pathological slide information and slide recognition results of historical slide scans;

[0143] Preprocessing the pathological section information and the section recognition results of the historical slide scans to obtain a training sample set;

[0144] Inputting the training sample set into the initialized pathology recognition neural network model for training;

[0145] Get the accuracy of the output results;

[0146] If the accuracy rate is greater than a preset accuracy rate threshold, the training is stopped to obtain the pathology recognition neural network model.

[0147] It should be noted that the pathology recognition neural network model requires a large amount of historical data for training. The larger the amount of data, the more accurate the result. The pathology recognition neural network model in this application can be trained by using the pathology slice information and slice recognition results of the historical slide scans as input. Of course, when training the neural network model, it is necessary not only to train with the pathology slice information and slice recognition results of the historical slide scans, but also to train in combination with the determined pathology slice disease causes and pathogenesis. By comparing a large amount of test data with real data, the results obtained will be more accurate, thereby making the output results of the pathology recognition neural network more accurate. Preferably, the accuracy threshold is generally set to 90%.

[0148] According to an embodiment of the present invention, it also includes counting the workload of the user terminal, specifically:

[0149] Defining the output of the recognition result of the pathological section once as a workload;

[0150] Counting and quantifying the total workload of the user terminal based on a preset period;

[0151] A corresponding remuneration mechanism is matched based on the total workload value to output a total remuneration value.

[0152] It should be noted that the workload statistics of the pathologist can also be obtained by counting the output results, counting the output results within a preset period (for example, 30 days), and quantifying them as the total workload of the user end to match the corresponding remuneration mechanism to output the total remuneration value. For example, the total workload value is "100" and the corresponding total remuneration value is "1500"; it is worth mentioning that only the output review and judgment result as the identification result of the pathological section can become a workload, thereby ensuring the work quality of the pathologist.

[0153] Furthermore, when allocating tasks, it also includes matching users according to task label information and completing task scheduling to improve the overall task processing efficiency and accuracy. Simply put, it is to allow more suitable users to complete the corresponding tasks. Among them, for the scheduling of processing time: the pathological slices are marked with time labels to obtain the urgency of the processing, and the priority tasks are scheduled accordingly. For example, users with faster average task response and shorter processing time can receive more tasks, and more urgent tasks will be distributed to such users first. For the scheduling of task matching, due to the different abilities and advantages of different pathologists, it can be judged based on past task processing results, and more suitable tasks can be recommended to doctors. For example, according to past results, doctor A has an accuracy rate of more than 80% in the interpretation of HSIL slices. Then, when AI processing considers suspected HSIL, the reading task of the slice will be distributed to doctor A for processing first.

[0154] It is worth mentioning that the pathological interpretation method based on remote film reading and annotation proposed in this application also includes grading the marked target suspicious area, specifically:

[0155] Acquire the target suspicious area of ​​the pathological section;

[0156] Counting the number of slice data matched by the target suspicious area;

[0157] The proportion of the number of slice data values ​​can be calculated based on the target area, wherein:

[0158] If the proportion exceeds the level threshold, the level of the target suspicious area is classified as the first level;

[0159] If the proportion does not exceed the level threshold, the level of the target suspicious area is classified as the second level.

[0160] It should be noted that the degree of change of the diseased cells in the target suspicious area of ​​the pathological section is different, so it is necessary to count the matching slice data in the target suspicious area to judge the degree of the lesion accordingly. The level threshold is taken as 60%. If the proportion exceeds 60%, it indicates that the degree of diseased cells in the target suspicious area has exceeded 60%. The level of the target suspicious area is divided into the first level. When the user-side pathologist interprets it, it can be interpreted according to different level divisions.

[0161] The third aspect of the present invention provides a computer-readable storage medium, which includes a machine-based pathology interpretation method program based on remote film reading and annotation. When the pathology interpretation method program based on remote film reading and annotation is executed by a processor, the steps of a pathology interpretation method based on remote film reading and annotation as described in any one of the above items are implemented.

[0162] The present invention discloses a pathology interpretation method, system and readable storage medium based on remote film reading and annotation. Double pathology interpretation is set to improve the accuracy of pathology interpretation and trace back to the specific operator. A neural network model is set to automatically identify areas with high confidence in pathology sections, thereby reducing the repetitive workload of pathologists and improving the efficiency of pathology identification.

[0163] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0164] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0165] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0166] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0167] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

Claims

1. A pathological interpretation method based on remote film reading and annotation. It is characterized in that The following steps are involved: Obtain electronic scans of slides and digitally display pathological section information to send to the user end for dual pathological interpretation; Inputting the pathological section information into a trained pathology recognition neural network model to obtain a simulated output result of the model; Acquire the target suspicious area of ​​the pathological slice based on the simulation output result, and synchronously share it with the user terminal; The two interpretation results returned by the user end are obtained and compared, and the interpretation results that are consistent with the comparison are output as the identification results of the pathological sections; the electronic scan of the glass slide is obtained, and the pathological section information is digitally displayed to be sent to the user end for double pathological interpretation, specifically: Scanning the glass slide with a preset scanner to obtain the electronic scanned copy; Matching the electronic scanned document based on a pre-stored slice database to obtain the pathological slice information, and digitally displaying the pathological slice information; The same pathological slice information is sent to the user terminal twice for pathological judgment, so as to obtain the test judgment result and the audit judgment result returned by the user terminal; the target suspicious area of ​​the pathological slice is obtained based on the simulation output result, and is synchronously shared with the user terminal, specifically: Extracting the simulation output results, calling the simulation output database to identify the pathological section analysis results; Obtaining a target suspicious area of ​​the pathological section based on the analysis result of the pathological section; The pathological slice attribute information is obtained for identity recognition, and the target suspicious area is synchronized with the pathological slice information of the corresponding identity for updating, so as to provide a reference for the user to perform pathological interpretation.

2. A pathology interpretation method based on remote film reading and annotation according to claim 1, It is characterized in that The pathology recognition neural network model training method is: Obtain pathological slide information and slide recognition results of historical slide scans; Preprocessing the pathological section information and the section recognition results of the historical slide scans to obtain a training sample set; Inputting the training sample set into the initialized pathology recognition neural network model for training; Get the accuracy of the output results; If the accuracy rate is greater than a preset accuracy rate threshold, the training is stopped to obtain the pathology recognition neural network model.

3. A pathology interpretation method based on remote film reading and annotation according to claim 1, It is characterized in that The obtaining and comparing the two interpretation results returned by the user terminal, and outputting the interpretation results that are consistent with each other as the recognition results of the pathological sections, specifically includes: Obtaining the test interpretation result and the review interpretation result marked by the user; Calculate the offset of the two reading results, where: If the offset is less than or equal to a preset threshold, the comparison result is determined to be consistent; If the offset is greater than the preset threshold, the comparison result is determined to be inconsistent; The audit reading result that is compared and found to be consistent is output as the recognition result of the pathological section.

4. A pathology interpretation method based on remote film reading and annotation according to claim 1, It is characterized in that It also includes counting the workload of the user end, specifically: Defining the output of the recognition result of the pathological section once as a workload; Counting and quantifying the total workload of the user terminal based on a preset period; A corresponding remuneration mechanism is matched based on the total workload value to output a total remuneration value.

5. A pathology interpretation system based on remote film reading and annotation, It is characterized in that The invention comprises a memory and a processor, wherein the memory comprises a pathology interpretation method program based on remote film reading and annotation, and the pathology interpretation method program based on remote film reading and annotation is executed by the processor to implement the following steps: Obtain electronic scans of slides and digitally display pathological section information to send to the user for dual pathological interpretation; Inputting the pathological section information into a trained pathology recognition neural network model to obtain a simulated output result of the model; Acquire the target suspicious area of ​​the pathological slice based on the simulation output result, and synchronously share it with the user terminal; The two interpretation results returned by the user end are obtained and compared, and the interpretation results that are consistent with the comparison are output as the identification results of the pathological sections; the electronic scan of the glass slide is obtained, and the pathological section information is digitally displayed to be sent to the user end for double pathological interpretation, specifically: Scanning the glass slide with a preset scanner to obtain the electronic scanned copy; Matching the electronic scanned document based on a pre-stored slice database to obtain the pathological slice information, and digitally displaying the pathological slice information; The same pathological slice information is sent to the user terminal twice for pathological judgment, so as to obtain the test judgment result and the audit judgment result returned by the user terminal; the target suspicious area of ​​the pathological slice is obtained based on the simulation output result, and is synchronously shared with the user terminal, specifically: Extracting the simulation output results, calling the simulation output database to identify the pathological section analysis results; Obtaining a target suspicious area of ​​the pathological section based on the analysis result of the pathological section; The pathological slice attribute information is obtained for identity recognition, and the target suspicious area is synchronized with the pathological slice information of the corresponding identity for update, so as to provide a reference for the user to perform pathological interpretation.

6. A pathology interpretation system based on remote film reading and annotation according to claim 5, It is characterized in that The obtaining and comparing the two interpretation results returned by the user terminal, and outputting the interpretation results that are consistent with each other as the recognition results of the pathological sections, is specifically as follows: Obtaining the test interpretation result and the review interpretation result marked by the user; Calculate the offset of the two reading results, where: If the offset is less than or equal to a preset threshold, the comparison result is determined to be consistent; If the offset is greater than the preset threshold, the comparison result is determined to be inconsistent; The audit reading result that is compared and found to be consistent is output as the recognition result of the pathological section.

7. A computer-readable storage medium, It is characterized in that The computer-readable storage medium includes a pathology interpretation method program based on remote film reading and annotation. When the pathology interpretation method program based on remote film reading and annotation is executed by a processor, the steps of a pathology interpretation method based on remote film reading and annotation as described in any one of claims 1 to 4 are implemented.

Images