A disease diagnosis standard library construction method, device and equipment and storage medium

Abstract

Embodiments of the present application disclose a disease diagnosis standard library construction method, device and equipment and a storage medium. The method can obtain first historical medical data from a target disease sample library; first image feature data of the first historical medical data is extracted to obtain a first key feature training data set, first short-term memory parameters and first long-term memory parameters are obtained based on the first key feature training data set and a detection standard model; second historical medical data is obtained; second image feature data of the second historical medical data is extracted to obtain a second key feature training data set, second short-term memory parameters and second long-term memory parameters are obtained based on the second key feature training data set and the detection standard model; target disease diagnosis standard data in a time dimension is obtained based on the short-term memory parameters and the long-term memory parameters, and a disease diagnosis standard library is constructed based on the target disease diagnosis standard data. The scheme can improve the accuracy of disease detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data processing and analysis, in particular to a construction method and device of a disease diagnosis standard library, equipment and a storage medium. BACKGROUND

[0002] With the continuous development of medical testing technology, more and more medical data are collected and recorded. The disease sample library, as an important medical database, contains a large amount of data and information related to disease categories. The data types include image data such as forward scattering signal pictures and scatter plots for measuring the size or complexity of cells, intensity, spectrum and wavelength data of detection fluorescence signals measured after specific fluorescent staining markers for evaluating the presence and expression level of cell surface markers or intracellular molecules, cell morphology maps, etc.; text data including typing, cell characteristics, cell cycle, parameter expression, immunophenotype in blood disease detection. These data can collectively constitute a complete flow cytometry blood disease detection report sample for assisting doctors in diagnosis and treatment decision-making.

[0003] In the process of research and practice of related technologies, the inventors of the present application found that the diagnosis standard of a disease type in a disease category is certain at the current time, but medical detection and diagnosis are usually affected by multiple factors such as technology development, detection method and detection equipment precision, resulting in dynamic adjustment of the diagnosis standard of a disease type over time. However, the medical detection results and diagnosis results of each disease category in the disease sample library do not dynamically adjust over time, and the corresponding disease diagnosis standard does not evolve autonomously over time, resulting in low accuracy of disease detection. SUMMARY

[0004] The embodiments of the present application provide a construction method, device, equipment and storage medium of a disease diagnosis standard library to solve the technical problem that with the development of medical testing technology over time, the medical detection results and diagnosis results of each disease category in the disease sample library do not dynamically adjust over time, and the corresponding disease diagnosis standard does not evolve autonomously over time, resulting in low accuracy of disease detection.

[0005] In a first aspect, the embodiments of the present application provide a construction method of a disease diagnosis standard library, comprising:

[0006] obtaining first historical medical data of a target disease category at a first historical time from a target disease sample library, the first historical medical data including first case sample image data and first case sample text data;

[0007] extract first image feature data of the first case sample image data, perform fusion processing on the first image feature data and the first case sample text data to obtain first key feature training data set, obtain first short-term memory parameter and first long-term memory parameter of the first historical time diagnosis standard based on the first key feature training data set and a preset detection standard model;

[0008] obtain second historical medical data of the target disease category at a second historical time from the target disease sample library, the second historical medical data including second case sample image data and second case sample text data, the second historical time being a time point after the first historical time;

[0009] extract second image feature data of the second case sample image data, perform fusion processing on the second image feature data and the second case sample text data to obtain second key feature training data set, obtain second short-term memory parameter and second long-term memory parameter of the second historical time diagnosis standard based on the second key feature training data set and the detection standard model;

[0010] obtain target disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical time to the second historical time based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter and the second long-term memory parameter, and construct a disease diagnosis standard library based on the target disease diagnosis standard data.

[0011] Optionally, the first image feature data of the first case sample image data includes:

[0012] perform preprocessing on the first case sample image data to obtain first target case sample image data;

[0013] obtain first image feature data of the first target case sample image data based on the first target case sample image data and a preset feature learning model.

[0014] Optionally, the fusion processing on the first image feature data and the first case sample text data to obtain the first key feature training data set includes:

[0015] perform preprocessing on the first case sample text data to obtain first target case sample text data;

[0016] perform combination processing on the first image feature data and the first target case sample text data to obtain the first key feature training data set.

[0017] Optionally, before extracting the first image feature data of the first case sample image data, the method further comprises:

[0018] From the target disease sample library, historical medical data of the target disease category at a third historical time is obtained, and the historical medical data at the third historical time comprises third case sample image data;

[0019] The third case sample image data is preprocessed to obtain target historical case sample image data;

[0020] The target historical case sample image data is segmented to obtain a historical case image training set, a historical case image verification set, and a historical case image test set;

[0021] The original feature learning model is trained based on the historical case image training set, the historical case image verification set, and the historical case image test set to obtain a preset feature learning model.

[0022] Optionally, before obtaining the first short-term memory parameter and the first long-term memory parameter of the first historical time diagnosis standard based on the first key feature training data set and the preset detection standard model, the method further comprises:

[0023] From the target disease sample library, historical medical data of the target disease category at a fourth historical time is obtained, and the historical medical data at the fourth historical time comprises fourth case sample image data and fourth case sample text data;

[0024] Image feature data of the fourth case sample image data is extracted based on the feature learning model to obtain third image feature data;

[0025] The third image feature data and the fourth case sample text data are combined to obtain a target key feature training data set;

[0026] The original long short-term memory network is trained based on the target key feature training data set to establish a preset detection standard model.

[0027] Optionally, based on the first key feature training data set and the preset detection standard model, obtaining the first short-term memory parameter and the first long-term memory parameter of the first historical time diagnosis standard comprises:

[0028] A first sample time corresponding to the first historical medical data and a first detection result corresponding to the first sample time are obtained;

[0029] The first short-term memory parameter and the first long-term memory parameter of the diagnosis standard at the first historical moment are obtained by establishing a weight relationship among the first key feature training data set, the first sample time and the first detection result through the detection standard model.

[0030] Optionally, based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter and the second long-term memory parameter, disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical moment to the second historical moment is obtained, including:

[0031] Based on the first short-term memory parameter and the second long-term memory parameter, a first association weight between the diagnosis standard corresponding to the first historical medical data and the first detection result is obtained.

[0032] Based on the second short-term memory parameter and the second long-term memory parameter, a second association weight between the diagnosis standard corresponding to the second historical medical data and the corresponding second detection result is obtained.

[0033] Based on the diagnosis standard corresponding to the first historical medical data, the diagnosis standard corresponding to the second historical medical data, the first association weight and the second association weight, disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical moment to the second historical moment is generated.

[0034] In a second aspect, an embodiment of the present application provides a disease diagnosis standard library construction device, including:

[0035] A first acquisition unit is configured to acquire first historical medical data of a target disease category at a first historical moment from a target disease sample library, wherein the first historical medical data includes first case sample image data and first case sample text data.

[0036] A first feature data extraction unit is configured to extract first image feature data of the first case sample image data, perform fusion processing on the first image feature data and the first case sample text data, obtain a first key feature training data set, and obtain a first short-term memory parameter and a first long-term memory parameter of a diagnosis standard at the first historical moment based on the first key feature training data set and a preset detection standard model.

[0037] A second acquisition unit is configured to acquire second historical medical data of the target disease category at a second historical moment from the target disease sample library, wherein the second historical medical data includes second case sample image data and second case sample text data, and the second historical moment is a time point after the first historical moment.

[0038] a second feature data extraction unit configured to extract second image feature data of the second case sample image data, perform fusion processing on the second image feature data and the second case sample text data, obtain second key feature training data set, and obtain second short-term memory parameter and second long-term memory parameter of the second historical time diagnosis standard based on the second key feature training data set and the detection standard model;

[0039] a construction unit configured to obtain target disease diagnosis standard data of autonomous evolution of the target disease category from the first historical time to the second historical time in a time dimension based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter and the second long-term memory parameter, and construct a disease diagnosis standard library based on the target disease diagnosis standard data.

[0040] In a third aspect, an embodiment of the present application further provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes steps in the construction method of the disease diagnosis standard library provided by any of the embodiments of the present application.

[0041] In a fourth aspect, an embodiment of the present application further provides a storage medium, which stores a plurality of instructions, and the instructions are loaded by a processor to execute steps in the construction method of the disease diagnosis standard library provided by any of the embodiments of the present application.

[0042] Compared with the prior art, the technical solution provided by the embodiments of the present application has the following advantages: the method provided by the embodiments of the present application can obtain first historical medical data of a target disease category at a first historical time from a target disease sample library, the first historical medical data including first case sample image data and first case sample text data; first image feature data of the first case sample image data is extracted, the first image feature data and the first case sample text data are fused to obtain first key feature training data set, and based on the first key feature training data set and a preset detection standard model, first short-term memory parameters and first long-term memory parameters of a first historical time diagnosis standard are obtained; second historical medical data of the target disease category at a second historical time is obtained from the target disease sample library, the second historical medical data including second case sample image data and second case sample text data, and the second historical time is a time point after the first historical time; second image feature data of the second case sample image data is extracted, the second image feature data and the second case sample text data are fused to obtain second key feature training data set, and based on the second key feature training data set and the detection standard model, second short-term memory parameters and second long-term memory parameters of a second historical time diagnosis standard are obtained; based on the first short-term memory parameters, the first long-term memory parameters, the second short-term memory parameters and the second long-term memory parameters, target disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical time to the second historical time is obtained, and a disease diagnosis standard library is constructed based on the target disease diagnosis standard data. The disease diagnosis standard library construction method proposed in the present application can extract features from historical medical data of different historical time target disease categories to obtain short-term memory parameters and long-term memory parameters of different historical time diagnosis standards, optimize the diagnosis standards of the same disease category based on these short-term memory parameters and long-term memory parameters, generate diagnosis standards of autonomous evolution of the target disease category in the time dimension, and improve the accuracy of disease detection. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0044] Figure 1 is a flowchart of the disease diagnosis standard library construction method provided by the embodiments of the present application;

[0045] Figure 2FIG. 1 is a schematic diagram of a disease diagnosis standard library construction device provided by an embodiment of the present application;

[0046] Figure 3 FIG. 2 is a structural schematic diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort fall within the scope of protection of the present application.

[0048] It should be understood that, when used in the specification and the appended claims, the terms "comprise" and "include" indicate the presence of described features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0049] It should also be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, unless otherwise clearly indicated by the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0050] It should be further understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations thereof.

[0051] As used in the present specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to a determination" or "in response to detecting" according to the context. Similarly, the phrase "if determined" or "if detected [a described condition or event]" can be interpreted according to the context as meaning "upon determining" or "in response to determining" or "upon detecting [a described condition or event]" or "in response to detecting [a described condition or event]".

[0052] In order to solve the technical problem that, with the development of medical testing technology over time, the medical test results and diagnosis results of each disease category in the disease sample library do not dynamically adjust over time, and the corresponding disease diagnosis standards do not evolve autonomously over time, resulting in low accuracy of disease detection, the present application provides a disease diagnosis standard library construction method, which can generate diagnosis standards of each disease category that evolve autonomously over time, and can improve the accuracy of disease detection.

[0053] Figure 1 A disease diagnosis standard library construction method is provided for an embodiment of the present application, and the disease diagnosis standard library construction method comprises:

[0054] S101, acquiring first historical medical data of a target disease category at a first historical time from a target disease sample library, the first historical medical data comprising first case sample image data and first case sample text data.

[0055] The target disease sample library refers to a medical database, which can collect and record a large amount of medical data of various disease categories, including a large amount of data and information related to different disease categories, and the data types can include image data and text data.

[0056] The target disease category includes at least one disease type, for example, the target disease category can be flow cytometry blood diseases, or the target disease category can be flow cytometry blood diseases and digestive system diseases, or the target disease category can be flow cytometry blood diseases, digestive system diseases and respiratory system diseases, etc.

[0057] The first historical time can be the time point when the first historical medical data is stored in the target disease sample library, and the first historical medical data refers to the medical data of the target disease category stored in the target disease sample library at the first historical time, such as clinical examination data, laboratory test data, cell images and other related data, etc., and the data types of the first historical medical data can include image data and text data.

[0058] The first case sample image data refers to the image data of the medical data of the target disease category in the target disease sample library, for example, forward scattering signal pictures and scatter plots for measuring the size or complexity of cells, intensity, spectrum and wavelength data of detection fluorescence signals measured after specific fluorescent staining markers for evaluating the presence and expression level of cell surface markers or cell molecules, cell morphology maps, etc.

[0059] The first case sample text data refers to the text data of the medical data of the target disease category in the target disease sample library, for example, typing, cell characteristics, cell cycle text, parameter expression, immunophenotype, etc. in various disease detection categories.

[0060] In an example, since the medical data in the existing target disease sample library is mainly stored and recorded in the form of conventional material documents, the data is not fully utilized, and through the embodiment of the present application, the "sleeping data" (i.e., the first historical medical data and the second historical medical data) can be obtained from the existing target disease sample library for big data analysis, and the diagnosis standard over time can be obtained. Sufficient medical test samples and related data are sufficient conditions for obtaining a relatively accurate diagnosis standard. The updated diagnosis standard has practical significance for improving the accuracy of disease detection. According to the above updated diagnosis standard, the detection result can be more comprehensively reflected, so as to improve the accuracy of disease detection.

[0061] S102, extract first image feature data of the first case sample image data, fuse the first image feature data and the first case sample text data, obtain the first key feature training data set, and obtain the first short-term memory parameter and the first long-term memory parameter of the first historical time diagnosis standard based on the first key feature training data set and the preset detection standard model.

[0062] The first image feature data refers to data expressing the image features of the first case sample image data. A convolutional neural network technology can be used to train a model to obtain a trained preset feature learning model, and then the feature data of the first case sample image data is extracted through the feature learning model to obtain the first image feature data.

[0063] The fusion processing refers to the combination of the first image feature data and the first case sample data to generate a new data set, i.e., to generate the first key feature training data set. The first key feature training data set includes image feature parameters, text feature parameters and corresponding sample times at the first historical time.

[0064] The preset detection standard model can be based on a long short-term memory (□Long Short Term Memory□, LSTM) to establish a deep learning and training model of the target disease category diagnosis standard to obtain a detection standard model meeting the expectations.

[0065] The first short-term memory parameter and the first long-term memory parameter can reflect the weight relationship between the long-term dependence of the sample detection index and the detection result over time.

[0066] In an embodiment, the step of "extracting the first image feature data of the first case sample image data" can include:

[0067] The first case sample image data is preprocessed to obtain the first target case sample image data.

[0068] Based on the first target case sample image data and the preset feature learning model, first image feature data of the first target case sample image data is obtained.

[0069] The preprocessing of the first case sample image data can include cleaning, denoising, standardization, vectorization and the like of the first case sample image data, so as to ensure the accuracy and consistency of the data.

[0070] In an example, the first target case sample image data can be input into the preset feature learning model, and first image feature data representing the image features of the first target case sample image data is output. Specifically, a convolutional neural network technology can be used to construct an autonomous learning model to obtain the preset feature learning model, which is used to extract effective feature representations from original data such as the first target case sample image data. The convolutional layer is used to extract local features of the image such as scatter distribution, spectral wavelength, cell morphology, etc. By applying multiple convolution kernels and filters to perform convolution operation on the input image, the edge, distribution feature, shape and other local information of the image are captured. Through the training process, different convolution kernels learn to respond to specific features, and a Tanh activation function is used for nonlinear transformation of data to increase feature effect. Then, the pooling layer operation is performed, and the maximum pooling (Max Pooling) and average pooling (Average Pooling) are used to select the maximum value and average value in the feature map through the sliding pooling window, so as to reduce the size of the data while retaining important feature information of the image and enhancing the robustness and generalization ability of the network. Cross-entropy loss function is used in the process to distinguish diseases such as blood disease types. Finally, the full connection layer is entered to classify and decide the high-level feature map that collects local features, and the high-dimensional feature map is flattened into a one-dimensional vector as the output data of the first target case sample image data, so as to obtain the first image feature data.

[0071] In an embodiment, the step of "fusing and processing the first image feature data and the first case sample text data to obtain the first key feature training data set" can include:

[0072] The first case sample text data is preprocessed to obtain first target case sample text data.

[0073] The first image feature data and the first target case sample text data are combined to obtain the first key feature training data set.

[0074] In an example, the first case sample text data collected is preprocessed, linearly transformed and discretely processed, etc., to reduce the difference between the data and make the data distribution tend to be normally distributed. Then, the detection time of the case sample corresponding to the first case sample text data is taken as a feature, the data is converted into serialized data, and a merged vector is formed together with the one-dimensional vector corresponding to the first image feature data obtained above, to form a first key feature training data set.

[0075] In an embodiment, before the step of "extracting the first image feature data of the first case sample image data", it can further include:

[0076] From the target disease sample library, historical medical data of the target disease category at a third historical time is obtained, and the historical medical data at the third historical time includes third case sample image data.

[0077] The third case sample image data is preprocessed to obtain target historical case sample image data.

[0078] The target historical case sample image data is segmented to obtain a historical case image training set, a historical case image validation set and a historical case image test set.

[0079] The original feature learning model is trained based on the historical case image training set, the historical case image validation set and the historical case image test set to obtain a preset feature learning model.

[0080] The third historical time can be the time point at which the historical medical data at the third historical time is stored in the target disease sample library, and the preprocessing of the third case sample image data can include cleaning, denoising, standardizing, vectorizing, etc. of the third case sample image data, to ensure the accuracy and consistency of the data.

[0081] The historical case image training set is used to train the original feature learning model, the historical case image validation set is used to verify and adjust the parameters of the original feature learning model, and the historical case image test set is used to test and evaluate the parameters of the original feature learning model. Finally, a trained preset feature learning model can be obtained.

[0082] In an example, by automatically or manually collecting detection sample data of hematopathy patients, including clinical examination data, laboratory test data, cell images and other related data, etc., then preprocessing the collected case sample image data, including data cleaning, denoising, standardization, vectorization, etc., to ensure the accuracy and consistency of the data, the data set obtained after preprocessing is divided into training set, validation set and test set, which are used for training model, adjusting parameters and evaluating model performance, a convolutional neural network technology is used to construct an autonomous learning model (i.e. original feature learning model), the original feature learning model is trained by the historical case image training set, the parameters of the original feature learning model are adjusted by the historical case image validation set, and the parameters of the original feature learning model are tested and evaluated by the historical case image test set, and finally the trained preset feature learning model is obtained.

[0083] In an embodiment, before the step of "obtaining the first short-term memory parameter and the first long-term memory parameter of the first historical time diagnosis detection standard based on the first key feature training data set and the preset detection standard model", it can also include:

[0084] From the target disease sample library, obtain historical medical data of the target disease category at a fourth historical time, the historical medical data at the fourth historical time including fourth case sample image data and fourth case sample text data;

[0085] Based on the feature learning model, extract image feature data of the fourth case sample image data to obtain third image feature data;

[0086] Combining the third image feature data and the fourth case sample text data, a target key feature training data set is obtained;

[0087] Based on the target key feature training data set, the original long short-term memory network is trained to establish a preset detection standard model.

[0088] Wherein, the fourth historical time can be the time point when the historical medical data at the fourth historical time is stored in the target disease sample library, and the preprocessing of the fourth case sample image data can include cleaning, denoising, standardization, vectorization, etc. of the fourth case sample image data, to ensure the accuracy and consistency of the data.

[0089] Wherein, the fourth case sample text data can be preprocessed, and the data obtained after preprocessing is combined with the third image feature data to obtain the target key feature training data set.

[0090] The original long short-term memory network can include a long short-term memory network (LSTM). By training the long short-term memory network with the target key feature training dataset, a trained long short-term memory network can be obtained, which is also the preset detection standard model.

[0091] In one example, a deep learning and training model for disease detection standards can be established based on a Long Short-Term Memory (LSTM) network to obtain a detection standard model that meets expectations. Feature weights are calculated using the detection results, main detection parameters, missing parameters, and detection time sequence as weights. During the process, an adjusted nonlinear activation function ReLU is used to train the LSTM neural network model to keep the model's convergence speed stable. Finally, the first short-term memory parameters and the first long-term memory parameters of the diagnostic standard at the first historical moment are obtained and stored in the database.

[0092] In one embodiment, the step "based on the first key feature training dataset and a preset detection standard model, obtain the first short-term memory parameter and the first long-term memory parameter of the first historical moment diagnostic test standard" may include:

[0093] Obtain the first sample time and the corresponding first test result corresponding to the first historical medical data;

[0094] By establishing the weight relationship between the first key feature training dataset, the first sample time, and the first detection result through the detection standard model, the first short-term memory parameter and the first long-term memory parameter of the first historical moment diagnostic test standard are obtained.

[0095] In one example, medical testing and diagnosis are often influenced by multiple factors, including technological advancements, the accuracy of testing methods and equipment, leading to dynamic adjustments in diagnostic criteria for a particular disease over time. Short-term memory parameters and long-term memory parameters represent the weighted relationships between the sample's testing indicators and results over long periods, aiming to uncover changes in weights and updates in correlations over extended time. A testing standard model, such as an LSTM network, can be used to establish the weighted relationship between the first key feature training dataset, time, and testing results; this yields the first short-term memory parameters and the first long-term memory parameters of the diagnostic criteria at the first historical moment.

[0096] S103. Obtain the second historical medical data of the target disease category at the second historical moment from the target disease sample library. The second historical medical data includes the image data of the second case sample and the text data of the second case sample. The second historical moment is the time point after the first historical moment.

[0097] The second historical time point can be the time point at which the second historical medical data is stored in the target disease sample library. The second historical medical data refers to medical data of the target disease category stored in the target disease sample library at the second historical time point, such as clinical examination data, laboratory test data, cell images, and other related data. The data types of the first historical medical data can include image data and text data.

[0098] The second case sample image data refers to image data of medical data of the target disease category in the target disease sample library. The storage time of the second case sample image data is different from that of the first case sample image data. The second case sample image data is image data stored in the target disease sample library at the first historical time point, and the first case sample image data is image data stored in the target disease sample library at the second historical time point.

[0099] The first case sample text data refers to text data of medical data of the target disease category in the target disease sample library. The first case sample text data is text data stored in the target disease sample library at the first historical time point, and the second case sample text data is text data stored in the target disease sample library at the second historical time point. The second historical time point is later than the first historical time point.

[0100] S104, extracting second image feature data of the second case sample image data, and performing fusion processing on the second image feature data and the second case sample text data to obtain a second key feature training data set. Based on the second key feature training data set and the detection standard model, the second short-term memory parameter and the second long-term memory parameter of the second historical time point diagnosis standard are obtained.

[0101] The second image feature data refers to data expressing image features of the second case sample image data. As in step S102, the first image feature data is extracted by the first image feature data method. Similarly, the feature learning model can be used to extract the feature data of the second case sample image data to obtain the second image feature data.

[0102] The second image feature data and the second case sample text data can be combined to obtain the second key feature training data set. Similar to the first key feature training data set, the second key feature training data set includes image feature parameters, text feature parameters, and corresponding sample times at the second historical time point.

[0103] The second short-term memory parameter and the second long-term memory parameter can reflect the long-term dependence of the weight relationship between the detection index and the detection result of the sample over time. By comparing the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter, and the second long-term memory parameter, the weight change and the correlation relationship in the time dimension can be obtained, and the mutual relationship can reflect the change of the diagnostic standard over time.

[0104] In an embodiment, the step of "extracting second image feature data of the second case sample image data" can include:

[0105] The second case sample image data is preprocessed to obtain second target case sample image data.

[0106] Based on the second target case sample image data and the feature learning model, the second image feature data of the second target case sample image data is obtained.

[0107] In an embodiment, the step of "fusing and processing the second image feature data and the second case sample text data to obtain the second key feature training data set" can include:

[0108] The second case sample text data is preprocessed to obtain second target case sample text data.

[0109] The second image feature data and the second target case sample text data are combined to obtain the second key feature training data set.

[0110] In an embodiment, the step of "obtaining the second short-term memory parameter and the second long-term memory parameter of the second historical time diagnostic standard based on the second key feature training data set and the detection standard model" can include:

[0111] The second sample time corresponding to the second historical medical data and the corresponding second detection result are obtained.

[0112] The weight relationship between the second key feature training data set, the second sample time, and the second detection result is established by the detection standard model, and the second short-term memory parameter and the second long-term memory parameter of the second historical time diagnostic standard are obtained.

[0113] S105, based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter, and the second long-term memory parameter, the target disease diagnostic standard data of the target disease class from the first historical time to the second historical time is obtained. The disease diagnosis standard library is constructed based on the target disease diagnosis standard data.

[0114] The target disease diagnosis standard data includes a target disease diagnosis standard, and the diagnosis standard over time is obtained by performing big data analysis on the first historical medical data of the first historical time and the second historical medical data of the second historical time.

[0115] In an embodiment, the step of "obtaining disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical time to the second historical time based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter, and the second long-term memory parameter" can include:

[0116] obtaining a first association weight between the diagnosis standard corresponding to the first historical medical data and the first detection result based on the first short-term memory parameter and the second long-term memory parameter;

[0117] obtaining a second association weight between the diagnosis standard corresponding to the second historical medical data and the corresponding second detection result based on the second short-term memory parameter and the second long-term memory parameter;

[0118] generating disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical time to the second historical time based on the diagnosis standard corresponding to the first historical medical data, the diagnosis standard corresponding to the second historical medical data, the first association weight, and the second association weight.

[0119] The target diagnosis standard and the weight relationship obtained by the present application can cover the relationship between all detection indexes and detection results. In reality, the detection indexes in a single sample detection usually do not cover all indexes. In limited indexes, the detection result can be more comprehensively reflected according to the updated detection standard, that is, the target disease diagnosis standard, so that the accuracy of disease detection can be improved.

[0120] As can be seen from the above, the present embodiment can perform feature extraction on historical medical data of different historical time target disease categories to obtain short-term memory parameters and long-term memory parameters of diagnosis standards of different historical times. The diagnosis standards of the same disease category are optimized based on these short-term memory parameters and long-term memory parameters, and the diagnosis standards of the target disease category in the time dimension are generated. The accuracy of disease detection can be improved.

[0121] In order to better implement the above method, accordingly, the present embodiment also provides a disease diagnosis standard library construction device. The disease diagnosis standard library construction device can be integrated in a computer device, for example, and the disease diagnosis standard library construction device can include a data acquisition module, a feature extraction module, a diagnosis standard optimization module, and a diagnosis standard database. Figure 2The disease diagnosis standard library construction device can include a first acquisition unit 201, a first feature data extraction unit 202, a second acquisition unit 203, a second feature data extraction unit 204, and a construction unit 205, as follows:

[0122] (1) The first acquisition unit 201;

[0123] The first acquisition unit 201 is configured to acquire first historical medical data of a target disease category at a first historical time from a target disease sample library, wherein the first historical medical data includes first case sample image data and first case sample text data.

[0124] (2) The first feature data extraction unit 202;

[0125] The first feature data extraction unit 202 is configured to extract first image feature data of the first case sample image data, fuse the first image feature data and the first case sample text data to obtain a first key feature training data set, and acquire first short-term memory parameters and first long-term memory parameters of the first historical time diagnosis standard based on the first key feature training data set and a preset detection standard model.

[0126] In an embodiment, the first feature data extraction unit 202 includes:

[0127] The first preprocessing sub-unit 2021 is configured to preprocess the first case sample image data to obtain first target case sample image data.

[0128] The first acquisition sub-unit 2022 is configured to acquire first image feature data of the first target case sample image data based on the first target case sample image data and a preset feature learning model.

[0129] In an embodiment, the first feature data extraction unit 202 includes:

[0130] The second preprocessing sub-unit 2023 is configured to preprocess the first case sample text data to obtain first target case sample text data.

[0131] The combination sub-unit 2024 is configured to combine the first image feature data and the first target case sample text data to obtain a first key feature training data set.

[0132] In an embodiment, the first feature data extraction unit 202 includes:

[0133] The second acquisition sub-unit 2025 is configured to acquire a first sample time corresponding to the first historical medical data and a first detection result corresponding thereto.

[0134] The establishing subunit 2026 is configured to establish a weight relationship among the first key feature training data set, the first sample time, and the first detection result by using the detection standard model, to obtain first short-term memory parameters and first long-term memory parameters of the first historical time diagnosis standard.

[0135] (3) The second acquisition unit 203;

[0136] The second acquisition unit 203 is configured to acquire second historical medical data of the target disease category at a second historical time from the target disease sample library, the second historical medical data including second case sample image data and second case sample text data, and the second historical time being a time point after the first historical time.

[0137] (4) The second feature data extraction unit 204;

[0138] The second feature data extraction unit 204 is configured to extract second image feature data of the second case sample image data, to perform fusion processing on the second image feature data and the second case sample text data, to obtain a second key feature training data set, and to acquire second short-term memory parameters and second long-term memory parameters of the second historical time diagnosis standard based on the second key feature training data set and the detection standard model.

[0139] (5) The construction unit 205;

[0140] The construction unit 205 is configured to obtain target disease diagnosis standard data of autonomous evolution of the target disease category in a time dimension from the first historical time to the second historical time based on the first short-term memory parameters, the first long-term memory parameters, the second short-term memory parameters, and the second long-term memory parameters, and to construct a disease diagnosis standard library based on the target disease diagnosis standard data.

[0141] In an embodiment, the construction unit 205 includes:

[0142] The third acquisition subunit 2051 is configured to acquire a first association weight between a diagnosis standard corresponding to the first historical medical data and the first detection result based on the first short-term memory parameters and the second long-term memory parameters.

[0143] The third acquisition subunit 2052 is configured to acquire a second association weight between a diagnosis standard corresponding to the second historical medical data and a corresponding second detection result based on the second short-term memory parameters and the second long-term memory parameters.

[0144] The data generation subunit 2053 is configured to generate disease diagnosis standard data of autonomous evolution of the target disease category in a time dimension from the first historical time to the second historical time based on the diagnosis standard corresponding to the first historical medical data, the diagnosis standard corresponding to the second historical medical data, the first correlation weight, and the second correlation weight.

[0145] In an embodiment, the disease diagnosis standard library construction apparatus further includes:

[0146] The third acquisition unit 206 is configured to acquire historical medical data of the target disease category at a third historical time from the target disease sample library, and the historical medical data at the third historical time includes third case sample image data.

[0147] The preprocessing unit 207 is configured to pre-process the third case sample image data to obtain target historical case sample image data.

[0148] The segmentation unit 208 is configured to segment the target historical case sample image data to obtain a historical case image training set, a historical case image verification set, and a historical case image test set.

[0149] The first training unit 209 is configured to train and process an original feature learning model based on the historical case image training set, the historical case image verification set, and the historical case image test set to obtain a preset feature learning model.

[0150] In an embodiment, the disease diagnosis standard library construction apparatus further includes:

[0151] The fourth acquisition unit 210 is configured to acquire historical medical data of the target disease category at a fourth historical time from the target disease sample library, and the historical medical data at the fourth historical time includes fourth case sample image data and fourth case sample text data.

[0152] The third feature data extraction unit 211 is configured to extract image feature data of the fourth case sample image data based on the feature learning model to obtain third image feature data.

[0153] The combination unit 212 is configured to combine the third image feature data and the fourth case sample text data to obtain a target key feature training data set.

[0154] The second training unit 213 is configured to train an original long short-term memory network based on the target key feature training data set to establish a preset detection standard model.

[0155] In practice, the above various units can be implemented as independent entities, or combined as the same or several entities, and the specific implementation of the above various units can refer to the method embodiments above, which will not be repeated here.

[0156] As can be seen from the above, in the disease diagnosis standard library construction device, the first acquisition unit 01 acquires first historical medical data of a target disease category at a first historical time from a target disease sample library, the first historical medical data including first case sample image data and first case sample text data; then, the first feature data extraction unit 202 extracts first image feature data of the first case sample image data, performs fusion processing on the first image feature data and the first case sample text data to obtain first key feature training data set, and acquires first short-term memory parameters and first long-term memory parameters of the diagnosis standard at the first historical time based on the first key feature training data set and a preset detection standard model; the second acquisition unit 203 acquires second historical medical data of the target disease category at a second historical time from the target disease sample library, the second historical medical data including second case sample image data and second case sample text data, and the second historical time being a time point after the first historical time; the second feature data extraction unit 204 extracts second image feature data of the second case sample image data, performs fusion processing on the second image feature data and the second case sample text data to obtain second key feature training data set, and acquires second short-term memory parameters and second long-term memory parameters of the diagnosis standard at the second historical time based on the second key feature training data set and the detection standard model; and the construction unit 205 obtains target disease diagnosis standard data of autonomous evolution of the target disease category in the time dimension from the first historical time to the second historical time based on the first short-term memory parameters, the first long-term memory parameters, the second short-term memory parameters and the second long-term memory parameters, and constructs a disease diagnosis standard library based on the target disease diagnosis standard data.

[0157] The scheme can perform feature extraction on historical medical data of a target disease category at different historical times to obtain short-term memory parameters and long-term memory parameters of diagnosis standards at different historical times, optimize the diagnosis standards of the same disease category based on these short-term memory parameters and long-term memory parameters, and generate diagnosis standards of autonomous evolution of the target disease category in the time dimension, which can improve the accuracy of disease detection.

[0158] The disease diagnosis standard library construction device can be implemented in the form of a computer program, which can run on a computer device as shown in Figure 3 .

[0159] As Figure 3 shown, an embodiment of the present application provides a computer device, comprising a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 complete mutual communication through the communication bus 114,

[0160] The memory 113 is used for storing a computer program.

[0161] In an embodiment of the present application, the processor 111 is used for executing the program stored in the memory 113, and realizes the disease diagnosis standard library construction method provided by any one of the foregoing method embodiments.

[0162] Those skilled in the art can understand that all or part of the processes in the method of the above embodiment can be completed by a computer program instructing related hardware. The computer program can be stored in a storage medium, which is a computer readable storage medium. The computer program is executed by at least one processor in the computer system to realize the process steps of the above method embodiment.

[0163] Therefore, the present application further provides a storage medium. The storage medium can be a computer readable storage medium. The storage medium stores a computer program. The computer program is executed by a processor to make the processor execute the disease diagnosis standard library construction method provided by any one of the foregoing method embodiments.

[0164] The storage medium is an entity, non-transient storage medium, for example, can be a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk or an optical disk and various entity storage media which can store program codes. The computer readable storage medium can be non-volatile or volatile.

[0165] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in general in the above description. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0166] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. For example, the described apparatus embodiments are merely schematic. For example, the division of the units is merely a logical function division. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In a possible implementation process, the functions of the units can be implemented by using a process or a result of another process.

[0167] The steps in the method embodiments of the present application can be adjusted, combined and deleted according to actual needs. The units in the device embodiments of the present application can be combined, divided and deleted according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0168] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a terminal or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0169] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0170] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, these modifications and variations of the present application are intended to be included within the scope of the present application, and the claims of the present application and their equivalents. Therefore, the present application is intended to include these modifications and variations.

[0171] The above description is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for constructing a disease diagnostic standard library, characterized in that, include: Obtain first historical medical data of the target disease category at the first historical moment from the target disease sample database. The first historical medical data includes first case sample image data and first case sample text data. First image feature data is extracted from the first case sample image data. The first image feature data and the first case sample text data are fused to obtain a first key feature training dataset. Based on the first key feature training dataset and a preset detection standard model, the first short-term memory parameter and the first long-term memory parameter of the first historical moment diagnostic standard are obtained. Obtain second historical medical data of the target disease category at a second historical moment from the target disease sample database. The second historical medical data includes second case sample image data and second case sample text data. The second historical moment is a point in time after the first historical moment. The second image feature data of the second case sample image data is extracted, and the second image feature data and the second case sample text data are fused to obtain the second key feature training dataset. Based on the second key feature training dataset and the detection standard model, the second short-term memory parameter and the second long-term memory parameter of the second historical moment diagnostic standard are obtained. Based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter, and the second long-term memory parameter, the target disease diagnostic criteria data that autonomously evolves in the time dimension from the first historical moment to the second historical moment are obtained, and a disease diagnostic criteria library is constructed based on the target disease diagnostic criteria data.

2. The method according to claim 1, characterized in that, The first image feature data extracted from the first case sample image data includes: The first case sample image data is preprocessed to obtain the first target case sample image data; Based on the first target case sample image data and the preset feature learning model, the first image feature data of the first target case sample image data is obtained.

3. The method according to claim 2, characterized in that, The process of fusing the first image feature data and the first case sample text data to obtain the first key feature training dataset includes: The first case sample text data is preprocessed to obtain the first target case sample text data; The first image feature data and the first target case sample text data are combined and processed to obtain the first key feature training dataset.

4. The method according to claim 2, characterized in that, Before extracting the first image feature data from the first case sample image data, the method further includes: From the target disease sample database, obtain the historical medical data of the target disease category at the third historical moment, wherein the historical medical data at the third historical moment includes the image data of the third case sample; The image data of the third case sample is preprocessed to obtain the image data of the target historical case sample; The target historical case sample image data is segmented to obtain a historical case image training set, a historical case image validation set, and a historical case image test set. The original feature learning model is trained based on the historical case image training set, historical case image verification set, and historical case image test set to obtain the preset feature learning model.

5. The method according to claim 4, characterized in that, Before obtaining the first short-term memory parameter and the first long-term memory parameter of the first historical moment diagnostic standard based on the first key feature training dataset and the preset detection standard model, the method further includes: From the target disease sample database, obtain historical medical data of the target disease category at the fourth historical moment, wherein the historical medical data at the fourth historical moment includes image data and text data of the fourth case sample; Based on the feature learning model, image feature data of the fourth case sample image data is extracted to obtain the third image feature data; The third image feature data and the fourth case sample text data are combined and processed to obtain the target key feature training dataset; The original long short-term memory network is trained based on the target key feature training dataset to establish a preset detection standard model.

6. The method according to claim 5, characterized in that, The first short-term memory parameter and the first long-term memory parameter of the first historical moment diagnostic standard are obtained based on the first key feature training dataset and the preset detection standard model, including: Obtain the first sample time and the corresponding first test result corresponding to the first historical medical data; By establishing the weight relationship between the first key feature training dataset, the first sample time, and the first detection result through the detection standard model, the first short-term memory parameter and the first long-term memory parameter of the first historical moment diagnostic standard are obtained.

7. The method according to claim 6, characterized in that, The method of obtaining disease diagnostic criteria data for the target disease category that autonomously evolves over time from the first historical moment to the second historical moment, based on the first short-term memory parameter, the first long-term memory parameter, the second short-term memory parameter, and the second long-term memory parameter, includes: Based on the first short-term memory parameter and the second long-term memory parameter, obtain the first correlation weight between the diagnostic criteria corresponding to the first historical medical data and the first test result; Based on the second short-term memory parameter and the second long-term memory parameter, obtain the second correlation weight between the diagnostic criteria and the corresponding second test results corresponding to the second historical medical data; Based on the diagnostic criteria corresponding to the first historical medical data, the diagnostic criteria corresponding to the second historical medical data, the first association weight, and the second association weight, disease diagnostic criteria data for the target disease category are generated that evolve autonomously over the time dimension from the first historical moment to the second historical moment.

8. An apparatus for constructing a disease diagnostic criteria library, characterized in that, Includes a unit for performing the method as described in any one of claims 1-7.

9. A computer device, characterized in that, The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-7.

Images