Coronary artery blockage degree evaluation method and system based on magnetocardiogram

Abstract

The invention discloses a coronary artery blockage degree evaluation method based on magnetocardiogram, which comprises the following steps: acquiring a magnetocardiogram data set, and dividing the magnetocardiogram data set into a training set and a test set; giving a label of each case of magnetocardiogram data; training different classification models by adopting the training set and the labels and taking different coronary angiography blockage degrees as boundary lines to obtain trained classification models; the trained classification model is adopted to test the test set, and the positive probability of each magnetocardiogram data under different classification models and coronary angiography blocking degrees is obtained; performing weighted average on the positive probability of each classification model under different coronary angiography blockage degrees to obtain a score of each classification model; performing weighted average on scores of each magnetocardiogram data under different classification models to obtain a final score of each magnetocardiogram data; evaluating the coronary artery blockage percentage according to the final score; the invention further discloses a coronary artery blockage degree evaluation system based on the magnetocardiogram. According to the method, more accurate, objective and subdivided evaluation can be provided for the coronary artery blockage condition.

Description

Technical Field

[0001] The present invention relates to the field of magnetocardiogram image processing, and more specifically, to a method and system for evaluating coronary artery blockage based on magnetocardiogram. Background Art

[0002] Magnetocardiography (MCG) is a functional test that is more sensitive and reliable than conventional electrocardiography (ECG) because it can directly measure magnetic signals that are not affected by musculoskeletal factors. However, the interpretation of MCG reports usually requires experienced professionals, which limits its clinical application.

[0003] At present, magnetocardiography can usually only provide rough assessments such as mild, moderate, severe, etc., which is difficult to reflect the subtle differences in the severity of the disease. For some diseases that require more detailed assessments, the level classification is not accurate enough. In addition, this method is highly subjective and may be affected by the doctor's subjective judgment. Different doctors may classify the severity of the disease in the same patient at different levels.

[0004] The artificial intelligence (AI)-based magnetocardiography (MCG) model for assessing the severity of coronary artery disease (CAD) is currently a hot topic in this field. It can provide clinicians with an automatic and accurate diagnostic tool that may promote clinical acceptance. However, researchers have been working to improve the prediction accuracy of the binary classification model, using the MGG as a threshold to divide negative and positive according to a certain degree of blockage in coronary angiography, which is not accurate enough for assessing the severity of coronary artery disease. Summary of the invention

[0005] An object of the present invention is to provide a method and system for assessing the degree of coronary artery blockage based on magnetocardiography to at least solve the above-mentioned problems.

[0006] In order to achieve the purpose and other advantages of the present invention, the present invention provides a method for assessing the degree of coronary artery blockage based on magnetocardiogram, comprising: Obtain the magnetocardiogram dataset and randomly divide it into a training set and a test set; According to the coronary angiography results corresponding to each magnetocardiogram data in the training set, a label is given to each magnetocardiogram data according to the degree of coronary angiography blockage; Using the training set and its corresponding labels, different coronary angiography blockage degrees are used as boundaries to divide yin and yang, and binary classification training is performed on one or more classification models to obtain trained classification models under different coronary angiography blockage degrees; The trained classification model is used to test the test set to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees; The positive probability of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees is assigned a weight, and the weighted average is taken to obtain the score of each magnetocardiogram data under each classification model; Assign weights to the scores of each magnetocardiogram data under different classification models, and take the weighted average to obtain the final score of each magnetocardiogram data; The percentage of coronary artery occlusion was assessed according to the final score.

[0007] Preferably, the method further includes preprocessing each magnetocardiogram data in the magnetocardiogram data set, extracting magnetocardiogram parameters, and using the magnetocardiogram parameters as input of the classification model.

[0008] Preferably, the magnetocardiogram parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, and negative pole area change standard deviation.

[0009] Preferably, the different coronary angiography blockage degrees include 50% blockage, 70% blockage, 80% blockage, and 90% blockage.

[0010] Preferably, the classification model includes a machine learning model and / or a deep learning model.

[0011] Preferably, the machine learning model includes a random forest model and an extreme gradient boosting model.

[0012] Preferably, the deep learning model includes a long short-term memory network model and a deep residual network model.

[0013] The present invention also provides a system based on the above method, comprising: Get module, which is used for magnetocardiogram dataset and randomly divides it into training set and test set; A labeling module, which is used to label each magnetocardiogram data according to the coronary angiography result corresponding to each magnetocardiogram data in the training set according to the degree of coronary angiography blockage; A training module, which is used to use the training set and its corresponding labels to divide the positive and negative parts according to different coronary angiography blockage degrees, perform binary classification training on one or more classification models, and obtain trained classification models under different coronary angiography blockage degrees; A testing module, which is used to test the test set using the trained classification model to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees; A first data processing module is used to assign weights to the positive probabilities of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees, and perform weighted average to obtain the scores of each magnetocardiogram data under each classification model; The second data processing module is used to assign weights to the scores of each magnetocardiogram data under different classification models, and perform weighted average to obtain the final score of each magnetocardiogram data; An evaluation module is used to evaluate the coronary artery blockage percentage according to the final score.

[0014] The present invention also provides an electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the program includes a method for executing the above method.

[0015] The present invention also provides a computer-readable storage medium storing a computer program used in combination with an electronic device, wherein the computer program can be executed by a processor to implement the above method.

[0016] The present invention has at least the following beneficial effects: The present invention constructs a scoring mechanism to quantify the results of magnetocardiogram data, thereby providing a more accurate, objective and detailed assessment of coronary artery blockage, avoiding the influence of doctors' subjective judgment, allowing patients to understand the degree of coronary artery blockage more intuitively and promoting clinical acceptance more easily.

[0017] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of a method for assessing coronary artery blockage based on magnetocardiography according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with embodiments and drawings so that those skilled in the art can implement the invention with reference to the description.

[0020] It should be understood that the terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.

[0021] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0022] In one aspect of the present invention, Figure 1 As shown, a method for assessing the degree of coronary artery blockage based on magnetocardiography is provided, comprising: S1. Obtain the magnetocardiogram dataset and randomly divide it into a training set and a test set.

[0023] Specifically, the magnetocardiogram data of the patient is collected by a magnetocardiograph and the coronary angiography result of the patient is correspondingly obtained to form a magnetocardiogram data set.

[0024] S2. According to the coronary angiography results corresponding to each magnetocardiogram data in the training set, a label is given to each magnetocardiogram data according to the degree of coronary angiography blockage.

[0025] For example, if the coronary angiography result corresponding to the magnetocardiogram data is 55% blocked, then when the coronary angiography blockage degree of 50% is used as the boundary to divide yin and yang, the magnetocardiogram data is marked as positive; when the coronary angiography blockage degree is 70%, 80%, and 90% as the boundary to divide yin and yang, the magnetocardiogram data is marked as negative; if the coronary angiography result corresponding to the magnetocardiogram data is 75% blocked, then when the coronary angiography blockage degree is 50% and 70% as the boundary to divide yin and yang, The magnetocardiogram data of this example is marked as positive. When the degree of coronary angiography blockage is 80% and 90% as the boundary line for dividing yin and yang, the magnetocardiogram data of this example is marked as negative. If the coronary angiography result corresponding to the magnetocardiogram data is 85% blockage, then when the degree of coronary angiography blockage is 50%, 70%, and 80% as the boundary line for dividing yin and yang, the magnetocardiogram data of this example is marked as positive. When the degree of coronary angiography blockage is 90% as the boundary line for dividing yin and yang, the magnetocardiogram data of this example is marked as negative.

[0026] S3. Using the training set and its corresponding labels, the different coronary angiography blockage degrees are used as the boundary to divide yin and yang, and one or more classification models are trained for binary classification to obtain trained classification models under different coronary angiography blockage degrees.

[0027] Specifically, the coronary angiography blockage was m1%, m2%, ..., m n % (n is an integer) as the boundary to divide yin and yang, and use one or several machine learning models or deep learning models (referred to as models M1, M2, ..., M N , N is an integer) for binary classification training. At this time, the model trained with the coronary angiography blockage degree m1% as the boundary can be recorded as P 11 , P 21 ,…,P N1 , the model trained with m2% as the boundary is recorded as P 12 , P 22 ,…,P N2 ,…, the degree of coronary angiography blockage m nThe model trained with the bound is denoted as P 1n , P 2n ,…,P Nn , using the training set and its corresponding labels, respectively, the model P 11 , P 21 ,…,P N1 , P 12 , P 22 ,…,P N2 ,…,P 1n , P 2n ,…,P Nn Training is performed to obtain a trained classification model under different coronary angiography blockage levels.

[0028] S4. Use the trained classification model to test the test set to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees.

[0029] Exemplarily, the positive probability obtained by the classification model is recorded as p , the negative probability is recorded as q , p+q=1 .

[0030] S5. Assign weights to the positive probabilities of each classification model of each magnetocardiogram data under different degrees of coronary angiography blockage, and perform weighted average to obtain the scores of each magnetocardiogram data under each classification model.

[0031] Exemplarily, according to the hierarchical analysis method, a judgment matrix is ​​constructed to assign weights to the positive probability of each classification model corresponding to each magnetocardiogram data under different coronary angiography blockage degrees, and the positive probability is multiplied by the assigned weight as an additive score, thereby obtaining the score of each magnetocardiogram data under each classification model, which is recorded as: S i = w 1 q 1 +w 2 q 2+…+ w n q n (i≤N, an integer) in, S i For model M i The corresponding model score, q 1. q 2. … q n For model M i In coronary angiography, blockage m1%, m2%, ..., m n % corresponding to the positive probability,w 1. w 2. … w n is the weight corresponding to the positive probability.

[0032] S6. Assign weights to the scores of each magnetocardiogram data under different classification models, and perform weighted average to obtain the final score of each magnetocardiogram data.

[0033] Exemplarily, according to the hierarchical analysis method, a judgment matrix is ​​constructed to assign weights to the scores of each magnetocardiogram data under different classification models, and the scores are multiplied by the assigned weights as additive scores to obtain the final scores of each magnetocardiogram data, which are within 0-100 points and are recorded as: S =( W 1 S 1 +W 2 S 2+…+ W n S N )×100 in, S The final score for each magnetocardiogram data is S 1. S 2. … S N For models M1, M2, …, M N The corresponding model score, W 1. W 2. … W n is the weight of the corresponding model score.

[0034] S7. Evaluate the percentage of coronary artery occlusion based on the final score.

[0035] For example, if the final score of the magnetocardiogram data is 83, then the coronary artery blockage percentage of the patient corresponding to the magnetocardiogram data is 83%.

[0036] In this embodiment, the training set and its corresponding labels are first used to pre-train different classification models under different coronary artery blockage degrees to obtain trained classification models, and then the trained classification models are used to test the test set to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees, and then weights are assigned to the positive probability of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees, and weighted average is performed to obtain the score of each classification model, and finally weights are assigned to the scores of each magnetocardiogram data under different classification models, and weighted average is performed to obtain the final score of each magnetocardiogram data. The coronary artery blockage percentage can be evaluated through the final score, that is, this embodiment provides a coronary artery blockage scoring mechanism based on magnetocardiogram data. Compared with the existing classification model that can only judge the positive and negative under different coronary artery blockage degrees, the method of this embodiment can directly evaluate the coronary artery blockage percentage through the final score, which can provide a more accurate, objective and detailed evaluation of the coronary artery blockage situation, avoid the influence of the doctor's subjective judgment, make the patient understand the coronary artery blockage degree more intuitively, and promote clinical acceptance more easily.

[0037] In another embodiment, the method further includes preprocessing each magnetocardiogram data in the magnetocardiogram data set, extracting magnetocardiogram parameters, and using the magnetocardiogram parameters as input of the classification model.

[0038] Specifically, the preprocessing method and the parameter extraction method are both prior arts and will not be described in detail here. Exemplarily, the magnetic field diagram, current diagram, and butterfly diagram in the magnetocardiogram are first obtained according to the magnetocardiogram data, and then the parameters in the magnetic field diagram, current diagram, and butterfly diagram are extracted, and the parameters include but are not limited to magnetic field or current angle parameters and their change parameters in the TT segment, change parameters of magnetic pole distance in the TT segment, position change parameters of positive and negative poles in the TT segment, position change parameters of current in the TT segment, change parameters of positive and negative pole boundaries and areas in the TT segment, change parameters of positive and negative pole areas in the TT segment, positive and negative pole area ratio parameters, R peak to T peak amplitude ratio parameters, etc. The number of parameters does not affect the implementation of this method.

[0039] Furthermore, the magnetocardiogram parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, and negative pole area change standard deviation.

[0040] In another embodiment, the different coronary angiography blockage levels include 50% blockage, 70% blockage, 80% blockage, and 90% blockage.

[0041] In another embodiment, the classification model includes a machine learning model and / or a deep learning model.

[0042] Furthermore, the machine learning model includes a random forest model (RF) and an extreme gradient boosting model (XGB).

[0043] Furthermore, the deep learning model includes a long short-term memory network model (LSTM, Long Short-Term Memory) and a deep residual network model (ResNet, Residual Network).

[0044] Another aspect of the present invention further provides a system based on the above method, comprising: Get module, which is used for magnetocardiogram dataset and randomly divides it into training set and test set; A labeling module, which is used to label each magnetocardiogram data according to the coronary angiography result corresponding to each magnetocardiogram data in the training set according to the degree of coronary angiography blockage; A training module, which is used to use the training set and its corresponding labels to divide the positive and negative parts according to different coronary angiography blockage degrees, perform binary classification training on one or more classification models, and obtain trained classification models under different coronary angiography blockage degrees; A testing module, which is used to test the test set using the trained classification model to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees; A first data processing module is used to assign weights to the positive probabilities of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees, and perform weighted average to obtain the scores of each magnetocardiogram data under each classification model; The second data processing module is used to assign weights to the scores of each magnetocardiogram data under different classification models, and perform weighted average to obtain the final score of each magnetocardiogram data; An evaluation module is used to evaluate the coronary artery blockage percentage according to the final score.

[0045] In this embodiment, an acquisition module, a marking module, a training module, a testing module, a first data processing module, a second data processing module and an evaluation module are constructed through a software program. The training module uses a training set and its corresponding labels to pre-train different classification models under different coronary artery blockage degrees to obtain a trained classification model. The testing module uses the trained classification model to test the test set to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees. The first data processing module assigns weights to the positive probability of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees, performs weighted averaging, and obtains the score of each classification model. The second data processing module assigns weights to the scores of each magnetocardiogram data under different classification models, performs weighted averaging, and obtains the final score of each magnetocardiogram data. The evaluation module evaluates the coronary artery blockage percentage according to the final score, thereby providing a more accurate, objective and detailed evaluation of the coronary artery blockage situation, avoiding the influence of the doctor's subjective judgment, making the patient's understanding of the coronary artery blockage degree more intuitive, and easier to promote clinical acceptance.

[0046] On the other hand, the present invention also provides an electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the program includes a program for executing the above-mentioned method.

[0047] Exemplarily, the electronic device is a device including a processor (CPU / MCU / SOC) and a memory (ROM / RAM), such as a desktop computer, a portable computer, a smart phone, etc.

[0048] Another aspect of the present invention provides a computer-readable storage medium storing a computer program for use in combination with an electronic device, wherein the computer program can be executed by a processor to implement the above method.

[0049] Exemplarily, the computer-readable storage medium includes: ROM, RAM, magnetic disk, optical disk, and other media that can store program codes.

[0050] The present invention will be further described below with specific embodiments: Embodiment 1: A method for assessing the degree of coronary artery blockage based on magnetocardiography, comprising: (1) Obtaining the examination results of a certain number of patients by magnetocardiograph, and extracting parameters of the magnetic field diagram, current diagram, and butterfly diagram in the magnetocardiogram by using existing technology for later use, wherein the parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, and negative pole area change standard deviation; (2) Obtain the patient's coronary angiography results for future use; (3) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 50% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 50% division. 11 , Model P 21 ; (4) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 70% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 70% division. 12 , Model P 22 ; (5) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 80% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 80% division. 13 , Model P 23 ; (6) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 90% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 90% division. 14 , Model P 24 ; (7) Obtain the magnetocardiogram data to be tested and extract the magnetocardiogram parameters. Use the four prediction models P of the RF model. 11 , P 12 , P 13 , P 14 The prediction results of the RF model under different coronary angiography blockage degrees were obtained, including the positive probability and negative probability; the four prediction models P of the XGB model were used 21 , P 22 , P 23 , P 24 Predictions were made separately to obtain the prediction results of the XGB model under different coronary angiography blockage levels, including positive probability and negative probability; (8) According to the hierarchical analysis method, construct a judgment matrix for model P 11 , P 12 , P 13 , P 14Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data RF model; (9) According to the hierarchical analysis method, construct a judgment matrix for model P 21 , P 22 , P 23 , P 24 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data XGB model; (10) According to the hierarchical analysis method, a judgment matrix is ​​constructed to assign weights to the RF model and the XGB model to obtain the final score of the single case data.

[0051] The above evaluation method was used to perform magnetic heart scoring on 600 cases of magnetocardiogram data and perform correlation analysis with the corresponding coronary angiography results. The Pearson correlation coefficient was 0.81, and the result was greater than 0.8, indicating a strong correlation. In addition, according to the scoring results obtained by this evaluation method, the accuracy of dividing yin and yang at 50% was 95.7%, and the accuracy of 70% was 89.3%, 80% was 83.2%, and 90% was 82.5%; while the accuracy of the RF model was about 88%, 87%, 79%, and 80% when divided at 50%, 70%, 80%, and 90%, respectively, and the accuracy of the XGB model was about 90%, 85%, 80%, and 79%, respectively, which proved the effectiveness of this evaluation method.

[0052] Embodiment 2: A method for assessing the degree of coronary artery blockage based on magnetocardiography, comprising: (1) Obtaining the examination results of a certain number of patients by magnetocardiograph, and extracting parameters of magnetic field diagram, current diagram, and butterfly diagram in the magnetocardiogram by using existing technology for later use, wherein the parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, negative pole area change standard deviation, and RT angle; (2) Obtain the patient's coronary angiography results for future use; (3) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 30% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 30% division. 11 , Model P 21 ; (4) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 50% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 50% division. 12 , Model P 22 ; (5) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 70% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 70% division. 13 , Model P 23 ; (6) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 80% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 80% division. 14 , Model P 24 ; (7) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 90% angiography blockage. The RF model and XGB model were used for training to obtain the prediction model P based on the 90% division. 15 , Model P 25 ; (8) Obtain the magnetocardiogram data to be tested and extract the magnetocardiogram parameters. Use the five prediction models P of the RF model. 11 , P 12 , P 13 , P 14 , P 15 The prediction results of the RF model under different coronary angiography blockage degrees were obtained, including the positive probability and negative probability; the five prediction models P of the XGB model were used. 21 , P 22 , P 23 , P 24 , P 25 Predictions were made separately to obtain the prediction results of the XGB model under different coronary angiography blockage levels, including positive probability and negative probability; (9) According to the hierarchical analysis method, construct a judgment matrix for model P 11 , P 12 , P 13 , P 14 , P 15 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data RF model; (10) According to the hierarchical analysis method, the judgment matrix is ​​constructed for model P 21 , P 22 , P 23 , P 24 , P 25 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data XGB model; (11) According to the hierarchical analysis method, a judgment matrix is ​​constructed to assign weights to the RF model and the XGB model to obtain the final score of the single case data.

[0053] The above evaluation method was used to perform magnetocardiographic scoring on 600 magnetocardiographic data and perform correlation analysis with the corresponding coronary angiography results. The Pearson correlation coefficient was 0.85, and the result was greater than 0.8, indicating a strong correlation.

[0054] Embodiment 3: A method for assessing the degree of coronary artery blockage based on magnetocardiography, comprising: (1) Obtaining the examination results of a certain number of patients by magnetocardiograph, and extracting parameters in the magnetic field diagram, current diagram, and butterfly diagram of the magnetocardiogram by using existing technology for later use, wherein the parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, negative pole area change standard deviation, RT angle, sum of positive pole position changes, and sum of negative pole position changes; (2) Obtain the patient's coronary angiography results for future use; (3) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 30% angiography blockage. The RF model, XGB model, and LSTM network were used for training to obtain the prediction model P based on the 30% division. 11 , Model P 21 , Model P 31 ; (4) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 50% angiography blockage. The RF model, XGB model, and LSTM network were used for training to obtain the prediction model P based on the 50% division. 12 , Model P 22 , Model P 32 ; (5) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 70% angiography blockage. The RF model, XGB model, and LSTM network were used for training to obtain the prediction model P based on the 70% division. 13 , Model P 23 , Model P 33 ; (6) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 80% angiography blockage. The RF model, XGB model, and LSTM network were used for training to obtain the prediction model P based on the 80% division. 14 , Model P 24 , Model P 34 ; (7) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 90% angiography blockage. The RF model, XGB model, and LSTM network were used for training to obtain the prediction model P based on the 90% division. 15 , Model P 25 , Model P35 ; (8) Obtain the magnetocardiogram data to be tested and extract the magnetocardiogram parameters. Use the five prediction models P of the RF model. 11 , P 12 , P 13 , P 14 , P 15 The prediction results of the RF model under different coronary angiography blockage degrees were obtained, including the positive probability and negative probability; the five prediction models P of the XGB model were used. 21 , P 22 , P 23 , P 24 , P 25 The predictions were made separately to obtain the prediction results of the XGB model under different coronary angiography blockage degrees, including positive probability and negative probability; the five prediction models P using the LSTM network 31 , P 32 , P 33 , P 34 , P 35 Predictions were made separately to obtain the prediction results of the LSTM network under different coronary angiography blockage levels, including positive probability and negative probability; (9) According to the hierarchical analysis method, construct a judgment matrix for model P 11 , P 12 , P 13 , P 14 , P 15 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data RF model; (10) According to the hierarchical analysis method, the judgment matrix is ​​constructed for model P 21 , P 22 , P 23 , P 24 , P 25 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data XGB model; (11) According to the hierarchical analysis method, the judgment matrix is ​​constructed for model P 31 , P 32 , P 33 , P 34 , P 35 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data LSTM network; (12) According to the hierarchical analysis method, a judgment matrix is ​​constructed to assign weights to the RF model, XGB model, and LSTM network to obtain the final score of the single-case data.

[0055] The above evaluation method was used to perform magnetocardiographic scoring on 600 magnetocardiographic data and perform correlation analysis with the corresponding coronary angiography results. The Pearson correlation coefficient was 0.88, and the result was greater than 0.8, indicating a strong correlation.

[0056] Embodiment 4: A method for assessing the degree of coronary artery blockage based on magnetocardiography, comprising: (1) Obtaining the examination results of a certain number of patients by magnetocardiograph, and extracting parameters in the magnetic field diagram, current diagram, and butterfly diagram of the magnetocardiogram by using existing technology for later use, wherein the parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, negative pole area change standard deviation, RT angle, sum of positive pole position changes, and sum of negative pole position changes; (2) Obtain the patient's coronary angiography results for future use; (3) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 30% angiography blockage. The XGB model, KNN model, and ResNet network were used for training to obtain the prediction model P based on the 30% division. 11 , Model P 21 , Model P 31 ; (4) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 50% angiography blockage. The XGB model, KNN model, and ResNet network were used for training to obtain the prediction model P based on the 50% division. 12 , Model P 22 , Model P 32 ; (5) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 70% angiography blockage. The XGB model, KNN model, and ResNet network were used for training to obtain the prediction model P based on the 70% division. 13 , Model P 23 , Model P 33 ; (6) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 80% angiography blockage. The XGB model, KNN model, and ResNet network were used for training to obtain the prediction model P based on the 80% division. 14 , Model P 24 , Model P 34 ; (7) The coronary angiography results were used as the label of each patient's parameters, and the patients were divided into positive and negative according to the 90% angiography blockage. The XGB model, KNN model, and ResNet network were used for training to obtain the prediction model P based on the 90% division. 15 , Model P25 , Model P 35 ; (8) Obtain the magnetocardiogram data to be tested and extract the magnetocardiogram parameters. Use the five prediction models P of the XGB model. 11 , P 12 , P 13 , P 14 , P 15 The predictions were made separately to obtain the prediction results of the XGB model under different coronary angiography blockage degrees, including positive probability and negative probability; the five prediction models P using the KNN model 21 , P 22 , P 23 , P 24 , P 25 The predictions were made separately to obtain the prediction results of the KNN model under different coronary angiography blockage levels, including the positive probability and negative probability; the five prediction models P using the ResNet network 31 , P 32 , P 33 , P 34 , P 35 Predictions were made separately to obtain the prediction results of the ResNet network at different coronary angiography blockage levels, including positive probability and negative probability; (9) According to the hierarchical analysis method, construct a judgment matrix for model P 11 , P 12 , P 13 , P 14 , P 15 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data XGB model; (10) According to the hierarchical analysis method, the judgment matrix is ​​constructed for model P 21 , P 22 , P 23 , P 24 , P 25 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data KNN model; (11) According to the hierarchical analysis method, the judgment matrix is ​​constructed for model P 31 , P 32 , P 33 , P 34 , P 35 Divide the weights and multiply the positive probability by the assigned weight as the additive score to obtain the score of the single-case data ResNet network; (12) According to the hierarchical analysis method, a judgment matrix is ​​constructed to assign weights to the XGB model, KNN model, and ResNet network to obtain the final score of the single-case data.

[0057] The above evaluation method was used to perform magnetocardiographic scoring on 600 magnetocardiographic data and perform correlation analysis with the corresponding coronary angiography results. The Pearson correlation coefficient was 0.84, and the result was greater than 0.8, indicating a strong correlation.

[0058] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for assessing the degree of coronary artery blockage based on magnetocardiography, characterized in that: include: Obtain the magnetocardiogram dataset and randomly divide it into a training set and a test set; According to the coronary angiography results corresponding to each magnetocardiogram data in the training set, a label is given to each magnetocardiogram data according to the degree of coronary angiography blockage; Using the training set and its corresponding labels, different coronary angiography blockage degrees are used as boundaries to divide negative and positive, and binary classification training is performed on one or more classification models to obtain trained classification models under different coronary angiography blockage degrees; The trained classification model is used to test the test set to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees; The positive probability of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees is assigned a weight, and the weighted average is taken to obtain the score of each magnetocardiogram data under each classification model; Assign weights to the scores of each magnetocardiogram data under different classification models, and take the weighted average to obtain the final score of each magnetocardiogram data; The percentage of coronary artery occlusion was assessed according to the final score.

2. The method for assessing the degree of coronary artery blockage based on magnetocardiography according to claim 1, characterized in that: The method further includes preprocessing each magnetocardiogram data in the magnetocardiogram data set, extracting magnetocardiogram parameters, and using the magnetocardiogram parameters as input of the classification model.

3. The method for assessing the degree of coronary artery blockage based on magnetocardiography according to claim 2, characterized in that: The magnetocardiogram parameters include T peak magnetic field angle, T peak current angle, T peak maximum-minimum ratio, TT maximum current angle, TT maximum magnetic field angle, TT minimum current angle, TT minimum magnetic field angle, positive pole area change standard deviation, and negative pole area change standard deviation.

4. The method for assessing the degree of coronary artery blockage based on magnetocardiography according to claim 1, characterized in that: The different coronary angiography blockage degrees include 50% blockage, 70% blockage, 80% blockage, and 90% blockage.

5. The method for assessing the degree of coronary artery blockage based on magnetocardiography according to claim 1, characterized in that: The classification model includes a machine learning model and / or a deep learning model.

6. The method for assessing the degree of coronary artery blockage based on magnetocardiography according to claim 5, characterized in that: The machine learning models include random forest models and extreme gradient boosting models.

7. The method for assessing the degree of coronary artery blockage based on magnetocardiogram according to claim 5, characterized in that: The deep learning model includes a long short-term memory network model and a deep residual network model.

8. A system according to any one of claims 1 to 7, characterized in that: include: Acquisition module, which is used for the magnetocardiogram dataset and randomly divides it into training and test sets; A labeling module, which is used to label each magnetocardiogram data according to the coronary angiography result corresponding to each magnetocardiogram data in the training set according to the degree of coronary angiography blockage; A training module, which is used to use the training set and its corresponding labels to divide the positive and negative parts according to different coronary angiography blockage degrees, perform binary classification training on one or more classification models, and obtain trained classification models under different coronary angiography blockage degrees; A testing module, which is used to test the test set using the trained classification model to obtain the positive probability and negative probability of each magnetocardiogram data under different classification models and different coronary angiography blockage degrees; A first data processing module is used to assign weights to the positive probabilities of each classification model of each magnetocardiogram data under different coronary angiography blockage degrees, and perform weighted average to obtain the scores of each magnetocardiogram data under each classification model; The second data processing module is used to assign weights to the scores of each magnetocardiogram data under different classification models, and perform weighted average to obtain the final score of each magnetocardiogram data; An evaluation module is used to evaluate the coronary artery blockage percentage according to the final score.

9. An electronic device, characterized in that include: one or more processors; Memory; as well as One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising a method for executing any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that A computer program stored in a computer program used in combination with an electronic device, wherein the computer program can be executed by a processor to implement the method described in any one of claims 1 to 7.

Images