Automatic analysis and evaluation method for valvular stenosis based on AI identification and analysis

Abstract

The invention provides a valvular disease stenosis automatic analysis and evaluation method and device based on AI recognition and analysis and a storage medium. The method comprises the steps that cardiac image data and clinical data of a patient are collected; establishing an image recognition model, and obtaining valve leaflet key features by using the image recognition model; establishing a hemodynamic analysis model, and obtaining a valve stenosis degree index through the hemodynamic analysis model; based on the valve stenosis degree index, the stenosis degree of the valvular disease is automatically analyzed; and generating a valvular disease stenosis evaluation report based on the stenosis degree of the valvular disease. Cardiac images and clinical data of a patient are collected and preprocessed. An image recognition model is established to extract valve leaflet key features, and then a hemodynamic model is established to analyze the valve stenosis degree. And automatically analyzing the valvular disease stenosis degree and generating a report according to the stenosis degree index and the evaluation standard. Through accurate data processing and an advanced model, accurate valvular disease assessment is provided, the assessment time of a doctor is shortened, and the efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing and analysis, and particularly to an automated analysis and evaluation method, device, and storage medium for valvular stenosis based on AI recognition and analysis. Background Art

[0002] Cardiac valvular disease is a common cardiovascular disease, and valvular stenosis is one of the main causes leading to impaired cardiac function. Valvular stenosis is usually caused by abnormal valve structure (such as leaflet thickening, calcification, or deformation), resulting in limited valve opening and affecting the normal flow of blood. Traditional methods for evaluating valvular stenosis mainly rely on doctors' clinical experience, echocardiography, and analysis of hemodynamic parameters. However, these methods have problems such as strong subjectivity, time-consuming and laborious, and limited accuracy.

[0003] With the rapid development of artificial intelligence (AI) technology, especially the application of deep learning in image recognition and data analysis, new possibilities have been provided for the automated analysis and evaluation of valvular stenosis. AI technology can analyze a large amount of cardiac image data and clinical data to learn the relationship between the characteristics of valve structure and the degree of valvular stenosis and hemodynamic parameters, thereby realizing the automated evaluation of valvular stenosis.

[0004] Currently, although some AI-based methods for diagnosing valvular disease have been proposed, most of these methods focus on image recognition of valve structure, lacking in-depth analysis and comprehensive evaluation of hemodynamic parameters. In addition, these methods also have deficiencies in the quantitative evaluation of the degree of valvular stenosis and cannot provide accurate and reliable evaluation results for doctors. Summary of the Invention

[0005] The present invention aims to at least solve the technical problem in the prior art that there are also deficiencies in the quantitative evaluation of the degree of valvular stenosis and cannot provide accurate and reliable evaluation results for doctors, and particularly innovatively proposes an automated analysis and evaluation method, device, and storage medium for valvular stenosis based on AI recognition and analysis.

[0006] To achieve the above object of the present invention, the present invention provides an automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis, the method comprising:

[0007] S1. Collect the cardiac image data and clinical data of a patient, and preprocess the cardiac image data and clinical data;

[0008] S2. Establish an image recognition model, input the cardiac image data into the image recognition model, and use the image recognition model to obtain the key features of the valve leaflets;

[0009] S3. Establish a hemodynamic analysis model, input the preprocessed clinical data and the obtained key leaflet features into the hemodynamic analysis model, and obtain the valve stenosis degree index through the hemodynamic analysis model;

[0010] S4. Based on the valve stenosis degree index, combine with the preset stenosis degree evaluation criteria to automatically analyze the stenosis degree of the valvular disease;

[0011] S5. Generate a valvular disease stenosis evaluation report based on the stenosis degree of the valvular disease.

[0012] As an alternative embodiment of the present invention, optionally, in step S2, an image recognition model is established, and obtaining the key leaflet features by using the image recognition model includes:

[0013] S201. From the collected cardiac image data, select the image cardiac data containing clear valve structures as the training set and the validation set;

[0014] S202. Perform annotation and preprocessing on both the image cardiac data in the training set and the image cardiac data in the validation set;

[0015] S203. Use a convolutional neural network as the initial model, and use the labeled image cardiac data in the training set and data augmentation to train the initial model;

[0016] S204. Use the labeled image cardiac data in the validation set to verify the trained initial model, and iteratively optimize the initial model according to the verification results to obtain an image recognition model;

[0017] S205. Input the cardiac image data collected in step S1 into the image recognition model for leaflet feature extraction, and obtain the geometric shape features, thickness features, calcification degree features, and activity range features of the leaflets;

[0018] S206. Screen the key leaflet features based on the geometric shape features, thickness features, calcification degree features, and activity range features by using a feature importance evaluation method.

[0019] As an alternative embodiment of the present invention, optionally, establishing the hemodynamic analysis model in step S3 includes:

[0020] S301. Perform feature fusion on the preprocessed clinical data and the key leaflet features to obtain an input sequence data set;

[0021] S302. Use a regression model as the initial hemodynamic analysis model, and embed a blood flow model in the initial hemodynamic analysis model;

[0022] S303. Input the input sequence data set into the initial hemodynamic analysis model, train the initial hemodynamic analysis model, and add a regularization term during the training process of the initial hemodynamic analysis model;

[0023] S304. Use the cross-validation method to verify the trained initial hemodynamic analysis model, and iteratively optimize the initial hemodynamic analysis model according to the verification results to obtain a hemodynamic analysis model.

[0024] As an optional embodiment of the present invention, optionally, obtaining the valve stenosis degree index through the hemodynamic analysis model in step S3 includes:

[0025] S305. Initialize the hemodynamic analysis model;

[0026] S306. Standardize the clinical data and key leaflet features, and input the standardized clinical data and key leaflet features into the blood flow model;

[0027] S307. Based on the blood flow model, set the boundary conditions of blood flow according to the physiological structure and function of the heart;

[0028] S308. The blood flow model uses a numerical method to obtain the blood flow equation, and obtains the blood flow state in the heart based on the blood flow equation;

[0029] S309. Obtain the actual area of the valve orifice based on the flow state, compare the actual area of the valve orifice with the normal valve area, and obtain the valve area stenosis ratio;

[0030] S3010. Obtain the pressure difference across the valve and the blood flow through the valve based on the flow state;

[0031] S3011. Based on the valve area stenosis ratio, the pressure difference across the valve, and the blood flow through the valve, use the hemodynamic analysis model to obtain the valve stenosis degree index.

[0032] As an optional embodiment of the present invention, optionally, the clinical data includes hemodynamic parameters.

[0033] On the other hand, the present invention also provides a computer device, including:

[0034] A processor;

[0035] A memory for storing processor-executable instructions;

[0036] Wherein, when the processor is configured to execute the executable instructions, it implements the automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis.

[0037] On the other hand, the present invention also provides a computer-readable storage medium, including:

[0038] A memory having a computer program stored thereon;

[0039] A processor for executing the program in the memory to implement the automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis.

[0040] Advantages of the present invention: First, the present invention collects the cardiac image data and clinical data of patients and preprocesses them to ensure the accuracy and consistency of the data. Subsequently, an advanced image recognition model is used to extract features from the cardiac image data, especially for the key features of the valve leaflets, such as geometric shape, thickness, degree of calcification, and range of motion, etc. These features provide important input information for subsequent hemodynamic analysis. In the hemodynamic analysis stage, the present invention combines the clinical data and the key features of the valve leaflets, and constructs an input sequence data set through feature fusion technology. Then, a regression model is used as the initial model for hemodynamic analysis, and a blood flow model is embedded therein to simulate the blood flow situation in the heart. Through the training and iterative optimization of the model, the present invention can accurately obtain the valvular stenosis degree indicators, such as the stenosis ratio of the valve area, the pressure difference across the valve, and the blood flow through the valve, etc. Finally, based on these valvular stenosis degree indicators and combined with the preset stenosis degree evaluation criteria, the present invention can automatically analyze the stenosis degree of valvular disease and generate a detailed valvular stenosis evaluation report. The valvular stenosis evaluation report not only provides objective and accurate evaluation results for doctors, but also greatly shortens the evaluation time and improves work efficiency.

[0041] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. Description of the Drawings

[0042] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, wherein:

[0043] Figure 1 is a flowchart of the automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis of the present invention. Detailed Embodiments

[0044] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation of the present invention.

[0045] Embodiment 1

[0046] As Figure 1 shown, an automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis, the method comprising:

[0047] S1. Collect the cardiac image data and clinical data of the patient, and preprocess the cardiac image data and clinical data;

[0048] It should be noted that the preprocessing steps may include but are not limited to data cleaning, format conversion, denoising processing, etc., to ensure the accuracy and reliability of subsequent analysis. In the data cleaning stage, outliers can be removed, missing values can be filled, and the data can be standardized so that data from different sources can be compared on the same scale. Format conversion is to uniformly convert cardiac image data and clinical data in different formats into a format suitable for subsequent analysis. Denoising processing is to eliminate noise interference in the data and improve the signal-to-noise ratio of the data. Through this series of preprocessing steps, a high-quality data basis can be provided for subsequent analysis. In this embodiment, the clinical data includes hemodynamic parameters, electrocardiogram data, blood pressure data, heart rate data, etc., and these data can comprehensively reflect the cardiac function and status of the patient. When collecting these data, it is necessary to ensure the accuracy and integrity of the data for subsequent analysis and evaluation. For example, hemodynamic parameters can reflect the blood flow in the heart, including the pumping function of the heart, the resistance of blood vessels, etc.; electrocardiogram data can reflect the electrophysiological activities of the heart, including heart rate, heart rhythm, etc.; blood pressure data and heart rate data can reflect the pumping pressure and beating times of the heart and are important indicators for evaluating cardiac function. By comprehensively considering these clinical data, the stenosis degree of valvular disease can be more accurately evaluated and a more personalized treatment plan can be provided for the patient.

[0049] S2. Establish an image recognition model, input the cardiac image data into the image recognition model, and use the image recognition model to obtain key features of the valve leaf;

[0050] It should be noted that the specific process of establishing the image recognition model in step S2 includes key links such as data augmentation, model training, verification, and iterative optimization. The data augmentation technology effectively increases the diversity and quantity of training samples by performing a series of transformations on the initial cardiac image data, such as rotation, scaling, flipping, etc., which helps to improve the generalization ability of the model. Subsequently, the initial model is strictly verified using the labeled validation set data, and the model parameters are continuously adjusted according to the verification results. Through multiple iterations of optimization, an image recognition model with excellent performance is finally obtained. This model can accurately identify and extract the key features of the valve leaflets in the cardiac image, providing a solid foundation for subsequent hemodynamic analysis.

[0051] S3. Establish a hemodynamic analysis model, input the preprocessed clinical data and the obtained key features of the valve leaflets into the hemodynamic analysis model, and obtain the valve stenosis degree index through the hemodynamic analysis model;

[0052] It should be noted that the process of establishing the hemodynamic analysis model in step S3 fully considers the comprehensive influence of clinical data and key features of the valve leaflets. First, feature fusion is performed on the preprocessed clinical data and key features of the valve leaflets. This process combines different types of data organically through advanced feature fusion technology to form a more comprehensive and accurate input sequence data set. Subsequently, a regression model is used as the initial model for hemodynamic analysis, and a blood flow model is embedded in the model to simulate the complex blood flow situation in the heart. This embedded blood flow model can accurately simulate the blood flow state in the heart based on the physiological structure and function of the heart, providing key information for subsequent analysis.

[0053] In the model training stage, the input sequence data set is used to train the initial hemodynamic analysis model, and a regularization term is added during the training process to prevent the model from overfitting and improve the generalization ability of the model. After training, the cross-validation method is used to verify the model to ensure the accuracy and reliability of the model. According to the verification results, the model is iteratively optimized, and the model parameters are continuously adjusted to obtain the best hemodynamic analysis model. This model can accurately obtain the valve stenosis degree index, such as the stenosis ratio of the valve area, the pressure difference across the valve, and the blood flow through the valve, providing an important basis for evaluating the stenosis degree of valvular heart disease.

[0054] S4. Based on the valve stenosis degree index, combined with the preset stenosis degree evaluation criteria, automatically analyze the stenosis degree of valvular heart disease;

[0055] It should be noted that the preset stenosis degree evaluation criteria in step S4 are formulated based on a large amount of clinical data and hemodynamic studies, and can objectively reflect the severity of valvular stenosis. This standard usually includes multiple indicators, such as the stenosis ratio of the valve area, the pressure difference across the valve, the blood flow through the valve, etc. Each indicator has corresponding thresholds and evaluation grades. By comprehensively analyzing these indicators, the stenosis degree of valvular disease can be accurately judged, providing a personalized treatment plan for patients.

[0056] S5. Generate a valvular disease stenosis evaluation report based on the stenosis degree of the valvular disease.

[0057] It should be noted that the valvular disease stenosis evaluation report generated in step S5 details information such as the patient's valvular stenosis degree indicators, stenosis grade, possible causes, and recommended treatment plans. The report combines clear and intuitive charts and text descriptions, enabling doctors to quickly understand the patient's condition and formulate targeted treatment plans accordingly. In addition, this evaluation report can also serve as an important basis for patient condition tracking and treatment effect evaluation, providing strong support for doctor-patient communication.

[0058] In summary, the automatic analysis and evaluation method for valvular disease stenosis based on AI recognition and analysis in this embodiment first collects the patient's cardiac image data and clinical data and preprocesses them to ensure the accuracy and consistency of the data. Subsequently, an advanced image recognition model is used to extract features from the cardiac image data, especially for the key features of the valve leaflets, such as geometric shape, thickness, calcification degree, and range of motion. These features provide important input information for subsequent hemodynamic analysis. In the hemodynamic analysis stage, the present invention combines clinical data and key features of the valve leaflets to construct an input sequence data set through feature fusion technology. Then, a regression model is used as the initial model for hemodynamic analysis, and a blood flow model is embedded therein to simulate the blood flow in the heart. Through training and iterative optimization of the model, the present invention can accurately obtain valvular stenosis degree indicators, such as the stenosis ratio of the valve area, the pressure difference across the valve, and the blood flow through the valve. Finally, based on these valvular stenosis degree indicators and combined with the preset stenosis degree evaluation criteria, the present invention can automatically analyze the stenosis degree of valvular disease and generate a detailed valvular disease stenosis evaluation report. The valvular disease stenosis evaluation report not only provides objective and accurate evaluation results for doctors, but also greatly shortens the evaluation time and improves work efficiency.

[0059] As an alternative embodiment of the present invention, optionally, in step S2, an image recognition model is established, and using the image recognition model to obtain key features of the valve leaflets includes:

[0060] S201. Select the cardiac image data containing clear valve structures from the collected cardiac image data as the training set and the validation set;

[0061] It should be noted that in step S201, selecting the cardiac image data containing clear valve structures as the training set and the validation set is to ensure that the image recognition model can accurately learn and recognize the key features of the valve leaflets. These features are crucial for subsequent hemodynamic analysis because they directly reflect the structural and functional status of the valves. The training set data is used for the initial training of the model, while the validation set data is used to verify and adjust the model to ensure the accuracy and reliability of the model. When selecting data, relevant medical standards and specifications need to be strictly followed to ensure the accuracy and legality of the data. At the same time, the data also needs to be preprocessed, such as denoising and enhancement, to improve the learning efficiency and recognition accuracy of the model. Through this series of steps, an image recognition model with excellent performance can be established, providing a solid foundation for subsequent analysis and evaluation.

[0062] S202. Perform annotation and preprocessing on both the cardiac image data in the training set and the cardiac image data in the validation set;

[0063] It should be noted that in step S202, annotating the cardiac image data in the training set and the cardiac image data in the validation set is to provide clear learning objectives for the model, that is, to let the model know which key features need to be recognized and extracted. The annotation process is completed by professional medical imaging experts or doctors. They will perform detailed annotation on the image data according to the clinical characteristics and imaging manifestations of valvular diseases, including key features such as the geometric shape, thickness, calcification degree, and range of motion of the valve leaflets. These annotation information will serve as the supervision signal for model training, guiding the model to learn the correct feature extraction ability. At the same time, preprocessing the cardiac image data is also a crucial step. It includes operations such as denoising, enhancement, and standardization of the data to improve the quality and consistency of the data, which helps the model to better learn and generalize. Through this series of preprocessing and annotation steps, high-quality learning samples can be provided for subsequent model training, thus ensuring that the finally established image recognition model has excellent performance and accuracy.

[0064] S203. Use the convolutional neural network as the initial model, and use the annotated cardiac image data in the training set and data augmentation to train the initial model;

[0065] It should be noted that in step S203, the convolutional neural network (CNN) is selected as the initial model of the image recognition model because CNN has significant advantages in processing image data. CNN can automatically learn and extract features in images, especially spatial features, through structures such as convolutional layers, pooling layers, and fully connected layers. This is crucial for identifying the key features of valve leaflets in cardiac images. Using the labeled cardiac image data in the training set to train the initial model enables the model to learn the key features of the valve leaflets, such as geometric shape, thickness, calcification degree, and range of motion. At the same time, using data augmentation techniques to expand the training data can increase the generalization ability of the model, enabling it to better adapt to cardiac image data in different situations. During the training process, by continuously adjusting the model parameters and optimizing the model structure, an image recognition model with excellent performance can be obtained, providing a solid foundation for subsequent analysis and evaluation.

[0066] S204. Use the labeled cardiac image data in the validation set to validate the trained initial model, and iteratively optimize the initial model according to the validation results to obtain an image recognition model;

[0067] It should be noted that in step S204, using the labeled cardiac image data in the validation set to strictly validate the trained initial model is to ensure the accuracy and reliability of the model. The validation process evaluates the performance of the model by comparing the prediction results of the model with the actual annotation information. If there is a large deviation between the prediction results of the model and the actual annotation, the model needs to be adjusted and optimized. This process includes adjusting the model parameters, optimizing the model structure, etc., aiming to improve the recognition accuracy and generalization ability of the model. Through multiple iterations of optimization, an image recognition model with excellent performance is finally obtained. This model can accurately identify and extract the key features of valve leaflets in cardiac images, providing a solid foundation for subsequent hemodynamic analysis. In practical applications, this image recognition model can be combined with other analysis modules to form a complete automated analysis and evaluation system for valvular stenosis, providing a more accurate and efficient auxiliary diagnostic tool for doctors.

[0068] S205. Input the cardiac image data collected in step S1 into the image recognition model to extract valve leaflet features, and obtain the geometric shape features, thickness features, calcification degree features, and range of motion features of the valve leaflets;

[0069] It should be noted that in step S205, the preprocessed cardiac image data is input into the trained and optimized image recognition model, and the model is used to extract features from the cardiac image. The model can automatically identify and extract the geometric shape features of the valve leaflets, such as the opening and closing states of the valve leaflets, the contours and areas of the valve leaflets, etc.; at the same time, it also extracts the thickness features of the valve leaflets, that is, the thickness distribution and changes of the valve leaflets; in addition, the model also identifies the calcification degree features of the valve leaflets, including the location, scope, and degree of calcification, etc.; finally, the model also extracts the movement range features of the valve leaflets, that is, the movement trajectory and amplitude of the valve leaflets during the heartbeat process, etc. These feature information is crucial for subsequent hemodynamic analysis because they directly reflect the structure and function status of the valve, providing key information for subsequent analysis and evaluation. Through this step, the present invention can efficiently extract the key features of the valve leaflets in the cardiac image, laying a solid foundation for subsequent analysis and evaluation.

[0070] S206. Screen the key features of the valve leaflets by using the feature importance evaluation method based on the geometric shape features, thickness features, calcification degree features, and movement range features.

[0071] It should be noted that in step S206, based on the geometric shape features, thickness features, calcification degree features, and movement range features, the feature importance evaluation method is used to screen the extracted valve leaflet features to ensure the accuracy and effectiveness of subsequent hemodynamic analysis. The feature importance evaluation method quantifies the importance of features by calculating the contribution of each feature to the model prediction result. In this embodiment, various feature importance evaluation methods can be adopted, such as feature importance evaluation based on model weights, feature importance evaluation based on model output, and importance evaluation based on feature interaction effects, etc. By evaluating the importance of the extracted valve leaflet features, the key features that have the most influence on the evaluation of the valve stenosis degree can be screened out, providing more accurate and reliable information for subsequent analysis and evaluation. The implementation of this step further improves the accuracy and practicality in the automated analysis and evaluation of valvular stenosis.

[0072] As an optional embodiment of the present invention, optionally, the expression for extracting the valve leaflet features in step S205 is:

[0073]

[0074] A feature (i,j) = σ(Z feature (i,j))

[0075] P feature (i,j) = Pool(A feature (i′,j′)) for (i′,j′) ∈ pooling window

[0076] F feature = CNN(X, K feature , {σ i}, {Pool j )

[0077] Among them, Z feature (i, j) represents the convolution result of the cardiac image data at the position (i, j), i and j represent the position indices of the cardiac image data on the feature map, m and n represent the position indices on the convolution kernel, X(i + m, j + n) represents the value of the input cardiac image data at the position (i + m, j + n), k(m, n) represents the weight of the convolution kernel at the position (m, n), A feature (i, j) represents the value of the activated feature map at the position (i, j), σ() represents the activation function, P feature (i, j) represents the value of the pooled feature map at the position (i, j), Pool() represents the pooling operation, (i′, j′) represents the position index of (i, j) within the pooling window, pooling window represents the pooling window, F feature represents the leaflet feature, CNN() represents the operation of the convolutional neural network, X represents the input data, K feature represents the set of convolution kernels, {σ i} represents the set of activation functions, {Pool j} represents the set of pooling operations.

[0078] As an alternative embodiment of the present invention, optionally, establishing a hemodynamic analysis model in step S3 includes:

[0079] S301. Perform feature fusion on the preprocessed clinical data and the key leaflet features to obtain an input sequence dataset;

[0080] It should be noted that in step S301, feature fusion is a key step in organically combining clinical data and key leaflet features. The key leaflet features are extracted from cardiac image data and reflect the structural and functional status of the valve. By fusing clinical data and key leaflet features, a more comprehensive and accurate input sequence dataset can be constructed, providing richer and more valuable information for subsequent hemodynamic analysis. Feature fusion is to ensure that the fused dataset has high quality and consistency. After obtaining the input sequence dataset, the establishment and training of the hemodynamic analysis model can be carried out.

[0081] S302. Use a regression model as the initial hemodynamic analysis model and embed a blood flow model into the initial hemodynamic analysis model;

[0082] It should be noted that in step S302, the blood flow model is embedded to more accurately simulate the blood flow in the heart. The blood flow model is constructed based on the principles of fluid mechanics and physiology, and can simulate the blood flow state at the heart valves, including key parameters such as blood flow velocity and pressure distribution. By embedding the blood flow model into the regression model, the model can be used to more accurately simulate and analyze the blood flow in the heart. The implementation of this step enables the hemodynamic analysis model to better reflect the actual working state of the heart valves, providing a more accurate and reliable basis for subsequent analysis and evaluation. After embedding the blood flow model, the hemodynamic analysis model also needs to be trained and iteratively optimized to ensure its accuracy and reliability. In the initial hemodynamic analysis model, the blood flow model is embedded. The specific embedding method can be to use the parameters or outputs of the blood flow model as one of the input features of the regression model, or to combine the blood flow model with the regression model in parallel or in series. In this way, an organic combination of the hemodynamic analysis model and the blood flow model can be achieved, so as to more accurately simulate and analyze the blood flow in the heart. After embedding the blood flow model, the hemodynamic analysis model also needs to be further trained and iteratively optimized to improve the accuracy and reliability of the model. The specific training method can adopt optimization algorithms based on gradient descent, such as Stochastic Gradient Descent (SGD), Adam, etc., and minimize the error between the prediction result and the actual observation value by continuously adjusting the model parameters.

[0083] S303. Input the input sequence data set into the initial hemodynamic analysis model, train the initial hemodynamic analysis model, and add a regularization term during the training process of the initial hemodynamic analysis model;

[0084] It should be noted that in step S303, inputting the input sequence data set into the initial hemodynamic analysis model is to enable the model to learn the complex relationship between clinical data and key leaflet features, and how these features affect the hemodynamic state of the heart. By training the model, the model can be enabled to have the ability to predict the degree of valve stenosis. Adding a regularization term during the training process is to prevent the model from overfitting and improve the generalization ability of the model. The regularization term restricts the complexity of the model by constraining the model parameters, enabling it to better adapt to unseen data. Through this step, a hemodynamic analysis model with excellent performance can be obtained, providing a solid basis for subsequent analysis and evaluation. After training is completed, the model can be used to perform hemodynamic analysis on new cardiac imaging data and clinical data to evaluate the degree of valve stenosis.

[0085] S304. Validate the initially trained hemodynamic analysis model using the cross - validation method, and iteratively optimize the hemodynamic analysis model according to the validation results to obtain a hemodynamic analysis model.

[0086] It should be noted that in step S304, using the cross - validation method to strictly validate the initially trained hemodynamic analysis model is to ensure the stability and reliability of the model. The cross - validation method divides the data set into multiple subsets, which are used as the training set and the validation set respectively, to train and validate the model multiple times, so as to evaluate the performance of the model. If there is a large deviation between the prediction result of the model and the actual observation value, the model needs to be adjusted and optimized. This process includes adjusting model parameters, optimizing model structure, etc., aiming to improve the prediction accuracy and generalization ability of the model. Through multiple iterative optimizations, a hemodynamic analysis model with excellent performance is finally obtained. This model can accurately evaluate the hemodynamic state of the heart, especially the degree of valvular stenosis, and provide a more accurate and efficient auxiliary diagnostic tool for doctors. In practical applications, this hemodynamic analysis model can be combined with other analysis modules to form a complete automated analysis and evaluation system for valvular stenosis, providing strong support for the diagnosis and treatment of valvular diseases.

[0087] As an optional embodiment of the present invention, optionally, obtaining the valvular stenosis degree index through the hemodynamic analysis model in step S3 includes:

[0088] S305. Initialize the hemodynamic analysis model;

[0089] It should be noted that initializing the hemodynamic analysis model in step S305 is to ensure that the model is in a known state before starting the analysis. The initialization process includes setting the initial parameters of the model, configuring the initial structure of the model, etc. Through initialization, it can be ensured that each time the model is used, it starts from the same starting point, which is convenient for subsequent analysis and comparison. After initialization, the hemodynamic analysis model can start to receive the input sequence data set.

[0090] S306. Standardize the clinical data and the key features of the valve leaf, and input the standardized clinical data and the key features of the valve leaf into the blood flow model;

[0091] It should be noted that in step S306, the standardization of clinical data and key leaflet features is to eliminate the dimensional differences between different features and improve the training efficiency and prediction accuracy of the model. Standardization usually includes steps such as data decentralization and normalization, converting the data into the form of a standard normal distribution with a mean of 0 and a standard deviation of 1, or scaling the data to a specific range. By standardizing the clinical data and key leaflet features, it can be ensured that the model has the same scale when processing different features, thereby more accurately capturing the complex relationships between features. After the standardization process, the processed clinical data and key leaflet features are input into the blood flow model, and the model is used to simulate and analyze the blood flow in the heart, providing key information for the subsequent evaluation of the valve stenosis degree. The implementation of this step enables the hemodynamic analysis model to more accurately reflect the actual working state of the heart valve, providing a more accurate and reliable basis for subsequent analysis and evaluation.

[0092] S307. Set the boundary conditions of blood flow based on the physiological structure and function of the heart according to the blood flow model;

[0093] It should be noted that setting the boundary conditions of blood flow in step S307 is to ensure the accuracy and reliability of the blood flow model. The physiological structure and function of the heart are complex, including the four chambers of the heart, valves, blood vessels and other components, as well as their interactions. When setting the boundary conditions, it is necessary to fully consider the anatomical structure and physiological function of the heart, such as the opening and closing states of the valves, the contraction and relaxation of the myocardium, etc. These boundary conditions can affect the blood flow state in the heart, including key parameters such as blood flow velocity and pressure distribution. By reasonably setting the boundary conditions, the blood flow situation in the heart can be more accurately simulated, providing a more accurate and reliable basis for subsequent analysis and evaluation. In this embodiment, the boundary conditions include parameters such as the geometric shape of the heart, the movement law of the valves, and the contractility of the myocardium. These parameters can be obtained through various channels such as clinical data and cardiac imaging data and input into the blood flow model to achieve more precise simulation and analysis.

[0094] S308. The blood flow model uses a numerical method to obtain the blood flow equation and obtains the blood flow state in the heart based on the blood flow equation;

[0095] It should be noted that in step S308, the blood flow model uses a numerical method to solve the blood flow equation to accurately simulate and analyze the blood flow situation in the heart. The numerical method is a method of solving equations by discretizing continuous mathematical models. It can handle complex geometric shapes and boundary conditions and provide high-precision blood flow simulation results.

[0096] S309. Obtain the actual area of the valve orifice based on the flow state, compare the actual area of the valve orifice with the normal valve area, and obtain the stenosis ratio of the valve area;

[0097] It should be noted that in step S309, based on the flow state, the actual area of the valve orifice can be calculated, which reflects the effective flow cross-section of the valve when it is open. By comparing the actual area with the normal valve area, the stenosis ratio of the valve area can be calculated, which is an important indicator for evaluating the degree of valve stenosis. The higher the stenosis ratio, the more severely the flow capacity of the valve is limited, and the greater the impact on cardiac hemodynamics. The implementation of this step provides a direct quantitative indicator for the subsequent evaluation of the degree of valve stenosis, making the evaluation result more objective and accurate. After obtaining the stenosis ratio of the valve area, the degree of valve stenosis can be classified and evaluated according to this ratio, providing more clear and specific auxiliary diagnostic information for doctors.

[0098] S3010. Obtain the pressure difference across the valve and the blood flow through the valve based on the flow state;

[0099] It should be noted that in step S3010, based on the flow state, the pressure difference across the valve and the blood flow through the valve can be further analyzed. The pressure difference is one of the important indicators for evaluating valve function, which reflects the resistance of the valve to blood flow when it is open and closed. By measuring the pressure difference across the valve, it can be judged whether there are problems such as stenosis or regurgitation in the valve. At the same time, the blood flow through the valve is also an important basis for evaluating the degree of valve stenosis. The magnitude of the blood flow directly affects the pumping function of the heart and the hemodynamic state. When the valve is stenotic, the blood flow will decrease, resulting in the heart needing to increase its contractility to maintain sufficient blood circulation. Therefore, by measuring the blood flow through the valve, the impact of valve stenosis on cardiac function can be further understood. The implementation of this step provides more comprehensive information for the subsequent evaluation of the degree of valve stenosis, helping doctors more accurately judge the condition of patients with valvular heart disease and formulate reasonable treatment plans.

[0100] S3011. Based on the stenosis ratio of the valve area, the pressure difference across the valve, and the blood flow through the valve, use a hemodynamic analysis model to obtain an index of the degree of valve stenosis.

[0101] It should be noted that in step S3011, key information such as the stenosis ratio of the valve area, the pressure difference across the valve, and the blood flow rate through the valve is input into the hemodynamic analysis model. The model will comprehensively consider these factors and, through complex calculations and analyses, finally output an index of valve stenosis degree. This index is a quantitative value for comprehensively evaluating the degree of valve stenosis, which can objectively and accurately reflect the functional state and stenosis degree of the valve. Doctors can, based on this index, combined with the clinical data and cardiac imaging data of the patient, conduct a more comprehensive and accurate assessment of the condition of the valve disease patient, thereby providing strong support for formulating a reasonable treatment plan.

[0102] As an alternative embodiment of the present invention, optionally, in step S308, the expression of the blood flow equation is:

[0103]

[0104] where ρ represents the blood density, t represents time, v represents the blood velocity vector, represents the gradient operator, p represents the blood pressure, g represents the acceleration due to gravity, T represents the blood stress tensor, μ represents the blood viscosity coefficient, represents the transpose of the velocity gradient, and I represents the unit tensor.

[0105] As an alternative embodiment of the present invention, optionally, in step S3011, the expression for obtaining the valve stenosis degree index is:

[0106]

[0107] w1 + w2 + w3 = 1

[0108] where VSI represents the valve stenosis degree index, w1, w2, and w3 represent weight coefficients, A actual represents the actual area of the valve orifice, A normal represents the area of the normal valve, a, b, and c represent non-linear exponents, △p represents the pressure difference across the valve, p ref represents the reference pressure, Q represents the blood flow rate through the valve, Q ref represents the reference blood flow rate.

[0109] As an alternative embodiment of the present invention, optionally, the clinical data includes hemodynamic parameters.

[0110] Embodiment 2

[0111] A computer device, comprising:

[0112] A processor;

[0113] A memory for storing processor-executable instructions;

[0114] Wherein, when the processor is configured to execute the executable instructions, it implements the automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis in Embodiment 1.

[0115] It should be noted that the computer device includes: a processor and a memory. The computer device may further include one or more of a multimedia component, an input / output (I / O) interface, and a communication component.

[0116] The processor is used to control the overall operation of the computer device to complete all or part of the steps in the above-mentioned automated testing method for factory equipment based on big data.

[0117] The memory is used to store various types of data to support the operation of the computer device. These data may include, for example, instructions for any application or method operating on the computer device, as well as application-related data; the memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk.

[0118] The multimedia component may include a screen and an audio component. The screen may be a touch screen, for example, and the audio component is used to output and / or input audio signals; for example, the audio component may include a microphone, and the microphone is used to receive external audio signals. The received audio signals may be further stored in the memory or sent through the communication component; the audio component further includes at least one speaker for outputting audio signals.

[0119] The I / O interface provides an interface between the processor and other interface modules. The above-mentioned other interface modules may be a keyboard, a mouse, buttons, etc.; these buttons may be virtual buttons or physical buttons.

[0120] The communication component is used for wired or wireless communication between the computer device and other devices; wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G or 5G, or a combination of one or more of them. Accordingly, the communication component may include: a Wi-Fi module, a Bluetooth module, an NFC module, and a mobile communication module.

[0121] As a preferred solution of this embodiment, the computer device may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, and is used to execute the above-mentioned factory equipment automation test method based on big data.

[0122] Embodiment 3

[0123] A computer-readable storage medium, comprising:

[0124] A memory having a computer program stored thereon;

[0125] A processor for executing the program in the memory to implement the automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis in Embodiment 1.

[0126] It should be noted that the electronic device in the embodiments of the present disclosure includes a processor and a memory for storing processor-executable instructions. Among them, the processor is configured to implement the automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis described in any one of the foregoing when executing the executable instructions.

[0127] Here, it should be pointed out that the number of processors may be one or more. At the same time, in the electronic device of the embodiments of the present disclosure, an input device and an output device may also be included. Among them, the processor, the memory, the input device and the output device may be connected by a bus or in other ways, which is not specifically limited herein.

[0128] The memory, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and various modules, such as the programs or modules corresponding to an automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to an embodiment of the present disclosure. The processor executes various functional applications and data processing of the electronic device by running the software programs or modules stored in the memory.

[0129] The input device can be used to receive input numbers or signals. Among them, the signal can be a key signal related to the user settings and function control of the device / terminal / server. The output device can include display devices such as a display screen.

[0130] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and purposes of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. An automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis, characterized in that The method includes: S1. Collect the cardiac image data and clinical data of the patient, and preprocess the cardiac image data and clinical data; S2. Establish an image recognition model, input the cardiac image data into the image recognition model, and use the image recognition model to obtain the key features of the valve leaflets; S3. Establish a hemodynamic analysis model, input the preprocessed clinical data and the obtained key features of the valve leaflets into the hemodynamic analysis model, and obtain the valve stenosis degree index through the hemodynamic analysis model; S4. Based on the valve stenosis degree index, combined with the preset stenosis degree evaluation criteria, automatically analyze the stenosis degree of the valvular disease; S5. Generate a valvular disease stenosis evaluation report based on the stenosis degree of the valvular disease.

2. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 1, characterized in that, Establishing an image recognition model in step S2 and using the image recognition model to obtain the key features of the valve leaflets includes: S201. Select the image cardiac data containing clear valve structures from the collected cardiac image data as the training set and the validation set; S202. Perform annotation and preprocessing on both the image cardiac data in the training set and the image cardiac data in the validation set; S203. Use a convolutional neural network as the initial model, and use the annotated image cardiac data in the training set and data augmentation to train the initial model; S204. Use the annotated image cardiac data in the validation set to verify the trained initial model, and iteratively optimize the initial model according to the verification results to obtain an image recognition model; S205. Input the cardiac image data collected in step S1 into the image recognition model for valve leaflet feature extraction, and obtain the geometric shape features, thickness features, calcification degree features, and movement range features of the valve leaflets; S206. Screen the key features of the valve leaflets based on the geometric shape features, thickness features, calcification degree features, and movement range features using a feature importance evaluation method.

3. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 2, wherein, The expression for extracting the valve leaflet features in step S205 is: A feature (i,j) = σ(Z feature (i,j)) P feature (i,j) = Pool(A feature (i′,j′)) for (i′,j′) ∈ pooling window F feature = CNN(X, K feature , {σ i , {Pool j}) Among them, Z feature (i,j) represents the convolution result of the cardiac image data at the position (i,j). i and j represent the position indices of the cardiac image data on the feature map, m and n represent the position indices on the convolutional kernel, X(i+m,j+n) represents the value of the input cardiac image data at the position (i+m,j+n), k(m,n) represents the weight of the convolutional kernel at the position (m,n), A feature (i,j) represents the value of the activated feature map at the position (i,j), σ() represents the activation function, P feature (i,j) represents the value of the pooled feature map at the position (i,j), Pool() represents the pooling operation, (i′,j′) represents the position index of (i,j) within the pooling window, pooling window represents the pooling window, F feature represents the leaflet feature, CNN() represents the operation of the convolutional neural network, X represents the input data, K feature represents the set of convolutional kernels, {σ i} represents the set of activation functions, {Pool j} represents the set of pooling operations.

4. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 1, characterized in that, Establishing a hemodynamic analysis model in step S3 includes: S301. Perform feature fusion on the preprocessed clinical data and the key features of the valve leaflets to obtain an input sequence data set; S302. Use a regression model as the initial hemodynamic analysis model, and embed a blood flow model in the initial hemodynamic analysis model; S303. Input the input sequence data set into the initial hemodynamic analysis model, train the initial hemodynamic analysis model, and add a regularization term during the training process of the initial hemodynamic analysis model; S304. Use a cross-validation method to verify the trained initial hemodynamic analysis model, and iteratively optimize the initial hemodynamic analysis model according to the verification results to obtain a hemodynamic analysis model.

5. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 4, characterized in that Obtaining the valve stenosis degree index through the hemodynamic analysis model in step S3 includes: S305. Initialize the hemodynamic analysis model; S306. Standardize the clinical data and the key features of the valve leaflets, and input the standardized clinical data and the key features of the valve leaflets into the blood flow model; S307. Set boundary conditions for blood flow according to the physiological structure and function of the heart based on the blood flow model; S308. The blood flow model uses numerical methods to obtain the blood flow equation, and based on the blood flow equation, obtain the blood flow state in the heart; S309. Obtain the actual area of the valve orifice based on the flow state, and compare the actual area of the valve orifice with the normal valve area to obtain the stenosis ratio of the valve area; S3010. Obtain the pressure difference across the valve and the blood flow through the valve based on the flow state; S3011. Based on the stenosis ratio of the valve area, the pressure difference across the valve, and the blood flow through the valve, use the hemodynamic analysis model to obtain the valve stenosis degree index.

6. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 5, characterized in that, In step S308, the expression of the blood flow equation is: where ρ represents blood density, t represents time, v represents the velocity vector of blood, denotes the gradient operator, p represents blood pressure, g represents gravitational acceleration, T represents the stress tensor of blood, μ represents the viscosity coefficient of blood, represents the transpose of the velocity gradient, and I represents the unit tensor.

7. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 5, characterized in that In step S3011, the expression for obtaining the valve stenosis degree index is: w1 + w2 + w3 = 1 Among them, VSI represents the valve stenosis degree index, w1, w2, and w3 represent the weight coefficients, A actual represents the actual area of the valve orifice, A normal represents the area of the normal valve, a, b, and c represent the non-linear exponents, △p represents the pressure difference across the valve, p ref represents the reference pressure, Q represents the blood flow through the valve, Q ref represents the reference blood flow.

8. The automated analysis and evaluation method for valvular stenosis based on AI recognition and analysis according to claim 5 or 1, characterized in that, The clinical data includes hemodynamic parameters.

9. A computer device, characterized in that, Comprising: A processor; A memory for storing processor-executable instructions; Wherein, the processor is configured to implement the automated analysis and evaluation method for valve disease stenosis based on AI recognition and analysis according to any one of claims 1 to 8 when executing the executable instructions.

10. A computer-readable storage medium, characterized in that, Comprising: A memory having a computer program stored thereon; A processor for executing the program in the memory to implement the automated analysis and evaluation method for valve disease stenosis based on AI recognition and analysis according to any one of claims 1 to 8.

Images