Pathology detection system and method based on image recognition

Abstract

The invention relates to the technical field of pathology detection, in particular to a pathology detection system and method based on image recognition, and the method comprises the steps: obtaining pathology image data; key image feature values in the image copies are extracted to form an image feature vector set; constructing a pathological detection model I; the image data is processed through the first pathological detection model, and an analysis result is output; obtaining electronic medical record information from an electronic health record server, and obtaining medical record key information; a second pathological detection model is constructed, and a prediction result is output; the system is used for executing the method. The illness state is output through the first pathology detection model, the output result of the first pathology detection model is quantized and fused with the feature value extracted from the electronic medical record information, and the fused result is input into the second pathology detection model, so that whether a subject suffers from illness in the future or not can be predicted; whether the result of the first pathological detection model is accurate or not can be further detected, and the detection accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence dialogue technology, and in particular to a pathology detection system and method based on image recognition. Background Art

[0002] In pathology testing, tissue or cell samples must first be obtained from the patient. The preparation and quality of these samples directly affect the quality of subsequent images. Slices, stains (such as H&E staining, immunohistochemical staining, etc.) or other labeling processes are performed as needed. Images of the samples are captured using high-quality microscopes and cameras, which may include bright field, fluorescence, confocal and other microscopic imaging modes.

[0003] Ensuring high quality of sample preparation and accuracy of image acquisition is crucial for subsequent image analysis. With the development of artificial intelligence technology, especially the application of deep learning in the field of image processing, pathological diagnosis technology based on image recognition has made significant progress. However, there is a certain probability of error in the analysis of a single data source (i.e., image data), which may be due to factors such as image quality, sample preparation, staining differences, or limitations of the algorithm itself; how to reduce this error probability and improve the accuracy of detection is an urgent problem to be solved. Summary of the invention

[0004] Based on image recognition detection technology, the present invention combines an image detection system with a medical record system and fuses multiple data to further improve the accuracy of detection and predict the possibility of future illness to solve the problems in the background technology.

[0005] The technical solution provided by the present invention is: a pathology detection method based on image recognition, the method comprising: Acquiring pathological image data and preprocessing the image data to form an image copy that can be used for pathological analysis; sending a copy of the image to a storage module for storage; Extract key image feature values ​​from the image copy to form an image feature vector set; Constructing a pathology detection model 1 and training the pathology detection model 1; the output of the pathology detection model 1 is the disease state; The image data is processed through the trained pathology detection model and the analysis results are output; Obtain electronic medical record information from the electronic health record server and analyze it through natural language processing algorithms to obtain key medical record information; Construct pathology detection model 2, use the output value of pathology detection model 1 and key information of medical records as input variables of pathology detection model 2, and output the prediction results; The diseased state includes diseased and non-disease.

[0006] Preferably, the step of acquiring the pathological image data and preprocessing the image data to form an image copy that can be used for pathological analysis comprises the following steps: Acquiring pathological image data, wherein the pathological image data includes one or more of a tissue sample and a cell smear; Preprocess the image data to improve the image data quality; The preprocessing includes one or more of denoising, contrast enhancement, color correction, and image segmentation; The pre-processed image is backed up to form an image copy, which is sent to the storage module for storage.

[0007] Preferably, extracting key image feature values ​​from the image copy to form an image feature vector set comprises the following steps: Identify the shape of cells or tissues and obtain boundary length, boundary diameter, area, and eccentricity data; Identify the color of cells or tissues and obtain the area of ​​the corresponding color region in the image copy; Convert the image into a pixel grayscale space to obtain texture features, wherein the texture features include energy, contrast, and entropy; The area, eccentricity and texture features are combined into an image feature vector set.

[0008] Preferably, the step of forming an image feature vector set from area, eccentricity and texture features comprises the following steps: Construct the cell feature vector: , ;in They represent the normalized area eigenvalue, eccentricity eigenvalue, energy eigenvalue, and entropy eigenvalue respectively; Multiple cell feature vectors form a feature vector set , Indicates the number of cells.

[0009] Preferably, constructing a pathology detection model 1 and training the pathology detection model 1 comprises the following steps: Obtain historical image copies and analysis results from a historical database, extract key image feature values ​​from the historical image copies to form a feature vector set, and divide the analysis results and the feature vector set into a training set and a test set; The support vector machine model is used as the pathology detection model 1, and the RBF kernel is used as the kernel function of the pathology detection model 1; The training set is input into the model to train the model, the pathology detection model 1 is tested through the test set, and the hyperparameters of the pathology detection model 1 are optimized through cross-validation.

[0010] Preferably, the key information of the medical record obtained by analyzing through a natural language processing algorithm is: Extracting keywords from the electronic medical record to form a keyword set, extracting key feature values ​​from the keyword set to form a key feature value set, specifically includes the following steps: Clean the text data in the electronic medical record to remove blank characters and punctuation marks; Use NLP algorithms to standardize text data, including stemming and lemmatization; Use NLP algorithms to segment text data to facilitate keyword extraction; Build a keyword library containing professional medical terms; Combined with the keyword library, the keyword extraction algorithm is used to extract keywords from the text data, and the keywords are deduplicated and merged to form a keyword set; Extract key feature values ​​from the keyword set; The key eigenvalues ​​are quantized, and then the quantized key eigenvalues ​​are normalized to form a key eigenvalue set; the key eigenvalue set is: ,in is the number of key eigenvalues, .

[0011] Preferably, the construction of the second pathology detection model, using the output value of the first pathology detection model and the key information of the medical record as input variables of the second pathology detection model, and outputting the prediction result, comprises the following steps: A regression algorithm is used to construct a second pathology detection model, which is: ,in, is the intercept, is the input variable, is the output variable, is the regression coefficient; The diseased state output by the pathology detection model is quantified and set as , Take 1 or 0; Select two key eigenvalues ​​from the key eigenvalue set and Constitute the input variable set { },in ; Input the variables in the input variable set into the pathology detection model 2 and set the judgment threshold ,when When , it is predicted that the patient will not get sick in the future, otherwise, it is predicted that the patient will get sick; and the prediction result is output.

[0012] The present invention also provides a pathology detection system based on image recognition, the system is used to execute the pathology detection method based on image recognition, comprising: An image acquisition module, used for acquiring image data, wherein the image acquisition module comprises a microscope and / or a camera; An electronic health record server, used to store the electronic medical record information of the patient, the electronic health record server being connected to the intelligent processing module; Intelligent processing module, which processes and analyzes image data through machine learning algorithms, processes and analyzes electronic medical record information through natural language processing algorithms, and outputs analysis results; A display terminal, used to display the analysis results, the display terminal is connected to the intelligent processing module; Preferably, the intelligent processing module includes a processor, a storage module and a communication module, the storage module and the communication module are respectively connected to the processor, and the processor is communicatively connected to the electronic health record server through the communication module.

[0013] The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement a pathology detection method based on image recognition as described in any one of claims 1 to 7.

[0014] Beneficial effects of the present invention: 1. In the present invention, if the output value of pathology detection model 1 is 0 (indicating that the patient is not ill), pathology detection model 2 can use the key feature values ​​in the electronic medical record to predict the patient's risk of future illness. This can be achieved by analyzing the frequency of keywords and medical history records; if the output value of pathology detection model 1 is 1 (indicating that the patient is ill), medical record detection model 2 can combine pathology image analysis and key feature values ​​in the electronic medical record to further verify the accuracy of the diagnosis result. By combining image features and text features and performing multimodal analysis, the reliability of the diagnosis result can be improved.

[0015] 2. In the present invention, when processing the text data of the electronic medical record, the TF-IDF algorithm in the natural language processing technology is used to calculate the frequency (TF) and inverse document frequency (IDF) of each word in the text data, so as to obtain the weight of each word, thereby identifying the key words in the document; through the topic modeling algorithm, such as the latent Dirichlet distribution, the hidden topic structure in the document set can be identified. In this way, the text data in the electronic medical record can be classified according to the topic, and then the keywords related to the specific disease or pathological state can be extracted. Then, a professional vocabulary library containing important medical terms is constructed, including various diseases, treatment methods, drug names and other terms closely related to the medical field; through the professional vocabulary library, the NLP algorithm can more accurately identify and extract keywords related to the content of the electronic medical record, which helps to ensure that the extracted keywords are accurate and reliable, thereby providing high-quality input for subsequent data analysis and model training.

[0016] 3. In the present invention, by constructing a second pathology detection model, two key feature values ​​selected from the key feature value set are compared with the output value of the first pathology detection model. Constitute the input variable set { }, use the input variable set as the input variable of pathology detection model 2, output the prediction result, and the prediction result is the probability of future illness; set the judgment threshold ,when When the patient is predicted to be healthy, it is predicted that the patient will not be sick in the future; otherwise, it is predicted that the patient will be sick; and the prediction result is output, and the result can be displayed on a display terminal for the doctor to view. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of the method of the present invention; Figure 2 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0018] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be applied to other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.

[0019] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0020] Please combine Figure 1The present invention discloses a pathology detection method based on image recognition. The present invention provides a pathology detection method based on image recognition, the method comprising: Acquiring pathological image data and preprocessing the image data to form an image copy that can be used for pathological analysis; sending a copy of the image to a storage module for storage; Extract key image feature values ​​from the image copy to form an image feature vector set; Constructing a pathology detection model 1 and training the pathology detection model 1; the output of the pathology detection model 1 is a disease state, and the disease state includes diseased and not diseased; The image data is processed through the trained pathology detection model and the analysis results are output; Obtain electronic medical record information from the electronic health record server and analyze it through natural language processing algorithms to obtain key medical record information; Construct pathology detection model 2, use the output value of pathology detection model 1 and key information of medical records as input variables of pathology detection model 2, and output the prediction results to predict whether the examinee will become ill in the future.

[0021] First, pathology image data is obtained from various sources. The data may come from tissue samples, cell smears or other related biological samples. In order to improve the quality of the image data, a series of preprocessing operations are required. Filters or other algorithms can be used to remove random noise in the image to reduce its impact on subsequent analysis; the brightness and contrast of the image can be adjusted to make the features in the image more clearly visible. Ensure that the color of the image matches the color of the actual sample, which is particularly important for certain specific staining techniques; divide the image into different areas so that each area can be analyzed and processed separately. After completing the preprocessing, the image data is backed up to form image copies. These image copies are then sent to the storage module for storage for subsequent retrieval and use.

[0022] A copy of the image is obtained from the storage module and the boundaries of the cell or tissue are first determined by thresholding or the Canny algorithm. Once the boundaries are identified, it becomes straightforward to measure their length and diameter, which is usually done by counting the number of pixels on the boundary.

[0023] The area of ​​a cell or tissue can be obtained by counting the number of pixels within the boundary. The eccentricity is calculated by comparing the longest and shortest axes of the shape to show how much the cell or tissue shape deviates from a circle.

[0024] Then, the image is decomposed into basic color channels, such as RGB or HSV. For each channel, a color threshold is used to identify specific color regions, and then the areas of these regions are calculated. The color image is then converted to a grayscale image, specifically from the RGB color space to the grayscale space.

[0025] In the grayscale space, the gray level co-occurrence matrix (GLCM) is used to extract texture features, and features such as contrast, energy, and entropy are obtained through texture feature calculation.

[0026] All the above features (area, eccentricity, texture features) are integrated into a feature vector. The features are standardized or normalized to ensure that they are on the same scale, which is convenient for machine learning algorithms to learn.

[0027] The area, eccentricity and texture features are combined into an image feature vector set, which specifically includes the following steps: Construct the feature vector of the cell: , ;in They represent the normalized area eigenvalue, eccentricity eigenvalue, energy eigenvalue and entropy eigenvalue respectively; multiple eigenvectors constitute a eigenvector set ,in, Indicates the number of cells.

[0028] Obtain copies of historical images and their corresponding analysis results from the historical database. These data usually include the analysis results of the case by professional physicians and the image data related thereto.

[0029] Extract key image feature values ​​from historical image copies to form a feature vector set, which includes features such as shape, color, and texture.

[0030] The analysis results and feature vector set are divided into a training set and a test set. Usually, 80% of the data can be used as a training set and the remaining 20% ​​as a test set.

[0031] The support vector machine (SVM) model was used as the pathology detection model 1, and the radial basis function (RBF) was selected as the kernel function of the SVM.

[0032] The feature vectors of the training set and the corresponding analysis results are input into the SVM model to start training the model. This process will adjust the parameters of the SVM based on the provided training data in order to find the best decision boundary.

[0033] The trained pathology detection model was tested using the test set to evaluate the model's classification accuracy, recall rate and other performance indicators. The hyperparameters of the SVM were optimized through cross-validation.

[0034] For example, setting the regularization parameter C: For the regularization parameter C in SVM, a smaller value allows the model to accept more misclassifications, making the model smoother, but may lead to underfitting; while a larger value makes the model more sensitive to outliers in the data, which may cause overfitting. A common practice is to try exponential C values, such as [0.001, 0.01, 0.1, 1, 10,100, 1000].

[0035] Set the range of the RBF kernel parameter γ: a smaller γ value means a larger range of influence of the support vector, which may cause underfitting; a larger γ value means a smaller range of influence of the support vector, which may cause overfitting; a similar exponential series can also be used when searching for γ values, such as [0.001, 0.01, 0.1, 1, 10, 100]. GridSearch or Random Search can be used to find the optimal hyperparameter combination. Cross-validation can help us understand the generalization ability of the model for unknown data and avoid overfitting.

[0036] Through these steps, a SVM-based pathology detection model can be systematically trained and evaluated, enabling it to accurately classify and identify pathology images in practical applications.

[0037] The specific process of extracting keywords from electronic medical records through natural language processing (NLP) and forming a feature value set to facilitate the fusion of features in medical record information with image features is as follows: Cleaning text data in electronic medical records that may contain a large amount of extraneous whitespace characters and punctuation and correcting spelling errors; Standardize the text using NLP algorithms, including stemming (reducing words to their root form) and lemmatization (converting words to their base form). Medical terminology may vary by region, culture, or institution, so different terms need to be unified into standard terms for subsequent processing; Use NLP algorithms to segment text and divide sentences into separate vocabulary units to facilitate keyword extraction, because in medical terms, some words are composed of multiple words and need special processing to ensure that these words are not segmented incorrectly; Build a keyword library containing professional medical terms, and use keyword extraction algorithms (such as TF-IDF, graph-based algorithms, etc.) to extract keywords from text data based on the keyword library; Specifically: when the algorithm is processing, the words in the keyword library are used as important clues, and these words are extracted first or their weights are increased.

[0038] The extracted keywords are deduplicated and merged to form a clear keyword set for subsequent analysis. For example, "myocardial infarction" and "heart attack" can be merged into "cardiac event"; different word forms are normalized to the same form, such as unifying all "cancer" into "malignant tumor".

[0039] Extract key feature values ​​from the keyword set. These feature values ​​may include information such as the frequency of the word and the location of occurrence. Normalize the quantified key feature values ​​to ensure that they are on the same scale to facilitate subsequent data analysis and model training.

[0040] These keywords can also be further organized into a document-keyword matrix, where each document can be represented as a set of keywords, and then converted into a numerical feature vector, using machine learning algorithms to mine patterns in the data, which is beneficial for case classification and provides convenience for the next retrieval of medical record data.

[0041] The output of pathology detection model 1 is whether the person being tested has a certain disease. If the person is suffering from a certain disease, pathology detection model 2 can be used for further testing to determine whether the output result of pathology detection model 1 is accurate; if the person is not suffering from a certain disease, pathology detection model 2 can be used to predict whether the person will get sick in the future. The specific method is as follows: First, construct the pathology detection model 2, use the output value of the pathology detection model 1 and the key information of the medical record as the input variables of the pathology detection model 2, and output the prediction results, including the following steps: A regression algorithm is used to construct a second pathology detection model, which is: ,in, is the intercept, is the input variable, is the output variable, is the regression coefficient; The diseased state output by the pathology detection model is quantified and set as , Take 1 or 0. When the output of pathology detection model 1 is diseased, When 1 is taken, the output is not sick, Take 0; Select two key eigenvalues ​​from the key eigenvalue set and Constitute the input variable set { },in ; The timestamps and patient IDs corresponding to the two key feature values ​​and the output values ​​of case model one are kept consistent. The variables in the input variable set are input into pathology detection model two, and pathology detection model two outputs prediction results.

[0042] The selection of the two key feature values ​​can be based on the frequency of keywords extracted from the text analysis of the electronic medical record. For example, the frequency of occurrence of terms for certain diseases or symptoms (such as blood pressure, blood lipids, etc.) may be closely related to the patient's health status.

[0043] Please refer to Figure 2 The present invention also provides a pathology detection system based on image recognition, comprising: Image acquisition module, which contains a high-precision microscope and a high-resolution camera, is used to capture pathological sections or other medical images. These devices provide high-quality image data for subsequent analysis.

[0044] These devices themselves have data conversion functions, such as converting analog signals into digital signals, ensuring that image data can be effectively read and analyzed by the intelligent processing module.

[0045] Electronic health record server, which stores patients' electronic medical record information, including medical history, laboratory test results, surgical records, etc.

[0046] The electronic health record server needs to be connected to the intelligent processing module (wirelessly or wired) to ensure that the intelligent processing module can access and process the required medical record information in real time.

[0047] Intelligent processing module, the processor is responsible for executing machine learning and natural language processing algorithms to analyze image data and electronic medical record information.

[0048] The storage module is used to store algorithms, software, temporary data and processing results. It has sufficient storage space and fast data reading and writing capabilities. The communication module ensures that the intelligent processing module can exchange data with the electronic health record server and receive image data from the image acquisition module. The image data is analyzed through pre-trained machine learning algorithms to identify potential pathological changes; the natural language processing algorithm is used to extract key information from the electronic medical record to assist the diagnosis process.

[0049] Display terminal, the display terminal provides doctors or technicians with an intuitive user interface for displaying analysis results. This can include visualization of image analysis results, summaries of medical record information, etc. A touch display device can be selected as needed to facilitate users to enter queries, adjust analysis parameters, or select specific analysis functions.

[0050] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the pathology detection method based on image recognition.

[0051] The embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. The embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part, and / or installed from a removable medium. When the computer program is executed by the central processing unit (CPU), the above-mentioned functions defined in the method of the present application are executed. It should be noted that the above-mentioned computer-readable medium of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of an electrical, magnetic, optical, electromagnetic, infrared segment, or semiconductor, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wire segments, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, electrical wire, optical cable, RF, etc., or any suitable combination of the foregoing.

[0052] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0053] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments. Without departing from the principles, the implementation methods of the present invention may be deformed or modified in any way.

Claims

1. A pathology detection method based on image recognition, characterized in that: The method comprises: Acquiring pathological image data and preprocessing the image data to form an image copy that can be used for pathological analysis; sending a copy of the image to a storage module for storage; Extract key image feature values ​​from the image copy to form an image feature vector set; Constructing a pathology detection model 1 and training the pathology detection model 1; the output of the pathology detection model 1 is the disease state; The image data is processed through the trained pathology detection model and the analysis results are output; Obtain electronic medical record information from the electronic health record server and analyze it through natural language processing algorithms to obtain key medical record information; Construct pathology detection model 2, use the output value of pathology detection model 1 and key information of medical records as input variables of pathology detection model 2, and output the prediction results; The diseased state includes diseased and non-disease.

2. A pathology detection method based on image recognition according to claim 1, characterized in that: The method of obtaining the pathological image data and preprocessing the image data to form an image copy that can be used for pathological analysis includes the following steps: Acquiring pathological image data, wherein the pathological image data includes one or more of a tissue sample and a cell smear; Preprocess the image data to improve the image data quality; The preprocessing includes one or more of denoising, contrast enhancement, color correction, and image segmentation; The pre-processed image is backed up to form an image copy, which is sent to the storage module for storage.

3. The pathology detection method based on image recognition according to claim 1, characterized in that: The step of extracting key image feature values ​​from the image copy to form an image feature vector set includes the following steps: Identify the shape of cells or tissues and obtain boundary length, boundary diameter, area, and eccentricity data; Identify the color of cells or tissues and obtain the area of ​​the corresponding color region in the image copy; Convert the image into a pixel grayscale space to obtain texture features, wherein the texture features include energy, contrast, and entropy; The area, eccentricity and texture features are combined into an image feature vector set.

4. A pathology detection method based on image recognition according to claim 3, characterized in that: The step of combining area, eccentricity and texture features into an image feature vector set comprises the following steps: Construct the cell feature vector: , ;in They represent the normalized area eigenvalue, eccentricity eigenvalue, energy eigenvalue, and entropy eigenvalue respectively; Multiple cell feature vectors form a feature vector set , Indicates the number of cells.

5. The pathology detection method based on image recognition according to claim 1, characterized in that: Constructing a pathology detection model 1 and training the pathology detection model 1 includes the following steps: Obtain historical image copies and analysis results from a historical database, extract key image feature values ​​from the historical image copies to form a feature vector set, and divide the analysis results and the feature vector set into a training set and a test set; The support vector machine model is used as the pathology detection model 1, and the RBF kernel is used as the kernel function of the pathology detection model 1; The training set is input into the model to train the model, the pathology detection model 1 is tested through the test set, and the hyperparameters of the pathology detection model 1 are optimized through cross-validation.

6. A pathology detection method based on image recognition according to claim 1, characterized in that: The key information of the medical records obtained by analyzing through the natural language processing algorithm is: Extracting keywords from the electronic medical record to form a keyword set, extracting key feature values ​​from the keyword set to form a key feature value set, specifically includes the following steps: Clean the text data in the electronic medical record to remove blank characters and punctuation marks; Use NLP algorithms to standardize text data, including stemming and lemmatization; Use NLP algorithms to segment text data to facilitate keyword extraction; Build a keyword library containing professional medical terms; Combined with the keyword library, the keyword extraction algorithm is used to extract keywords from the text data, and the keywords are deduplicated and merged to form a keyword set; Extract key feature values ​​from the keyword set; The key eigenvalues ​​are quantized, and then the quantized key eigenvalues ​​are normalized to form a key eigenvalue set; the key eigenvalue set is: ,in is the number of key eigenvalues, .

7. A pathology detection method based on image recognition according to claim 6, characterized in that: The second pathology detection model is constructed, the output value of the first pathology detection model and the key information of the medical record are used as the input variables of the second pathology detection model, and the prediction result is output, including the following steps: A regression algorithm is used to construct a second pathology detection model, which is: ,in, is the intercept, is the input variable, is the output variable, is the regression coefficient; The diseased state output by the pathology detection model is quantified and set as , Take 1 or 0; Select two key eigenvalues ​​from the key eigenvalue set and Constitute the input variable set { },in ; Input the variables in the input variable set into the pathology detection model 2 and set the judgment threshold ,when , it is predicted that the patient will not get sick in the future; otherwise, it is predicted that the patient will get sick and the prediction result is output.

8. A pathology detection system based on image recognition, characterized in that: The system is used to execute a pathology detection method based on image recognition as described in any one of claims 1 to 7, comprising: An image acquisition module, used for acquiring image data, wherein the image acquisition module comprises a microscope and / or a camera; An electronic health record server, used to store the electronic medical record information of the patient, the electronic health record server being connected to the intelligent processing module; Intelligent processing module, which processes and analyzes image data through machine learning algorithms, processes and analyzes electronic medical record information through natural language processing algorithms, and outputs analysis results; The display terminal is used to display the analysis results, and the display terminal is connected to the intelligent processing module.

9. A pathology detection system based on image recognition according to claim 8, characterized in that: The intelligent processing module includes a processor, a storage module and a communication module. The storage module and the communication module are respectively connected to the processor, and the processor is communicatively connected to the electronic health record server through the communication module.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement a pathology detection method based on image recognition as described in any one of claims 1 to 7.

Images