Data analysis method and device based on magnetocardiogram

Abstract

The invention provides a data analysis method and device based on a magnetic cardiogram. The method comprises the following steps: acquiring magnetic cardiogram data containing a plurality of channels; carrying out preprocessing on the magnetic cardiogram data; extracting characteristic indexes from the preprocessed magnetic cardiogram data; based on the extracted characteristic indexes, constructing and training a machine learning model for predicting a coronary artery stenosis part; and analyzing the new magnetocardiogram data by utilizing a machine learning model so as to judge whether coronary artery diseases and narrow blood vessel parts exist or not. By extracting a plurality of characteristic indexes and combining with the machine learning model, the coronary artery stenosis part can be more accurately predicted, the accuracy of coronary artery disease diagnosis is improved, in addition, coronary artery angiography or coronary artery CT does not need to be carried out, and toxic and side effects of a contrast agent and an invasive operation can be prevented.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and particularly to a data analysis method and device based on magnetocardiogram. Background Art

[0002] Coronary Artery Disease (CAD) is one of the main causes of death and disability globally, seriously threatening human health. The diagnosis and treatment of CAD rely on accurate assessment methods, but existing diagnostic techniques (such as coronary angiography CAG, coronary computed tomography angiography CTA, cardiac magnetic resonance imaging CMR, positron emission tomography PET, and single photon emission computed tomography SPECT) have obvious technical limitations in clinical applications. These methods usually require expensive equipment and complex operation procedures, and some also have problems such as strong invasiveness, long examination time, and high radiation exposure, bringing great physical and psychological burdens to patients. In addition, the diagnostic sensitivity of these methods for early CAD is limited, making it difficult to meet the needs of precision medicine.

[0003] Magnetocardiography (MCG), as a non-invasive detection technique based on cardiac bio-magnetic field signals, has received extensive attention in recent years. MCG can provide highly sensitive cardiac electrophysiological information by recording the weak magnetic fields generated by cardiac electrical activities. However, in terms of technical applications, MCG data analysis has the following difficulties: First, the data dimension is high and complex. MCG data contains complex waveform information of multiple channels, and how to extract key feature indicators related to CAD from it is a technical difficulty. Second, there is a lack of mature analysis models. Existing research mostly stays at the data description level and it is difficult to accurately predict the vascular stenosis sites of CAD. How to achieve high-precision prediction of coronary artery stenosis sites through MCG data is an urgent problem to be solved. Summary of the Invention

[0004] In view of this, the present invention provides a data analysis method and device based on magnetocardiogram to solve at least one of the above-mentioned problems.

[0005] To achieve the above object, the present invention adopts the following solutions:

[0006] According to a first aspect of the present invention, there is provided a method for analyzing magnetocardiogram data, the method comprising: acquiring magnetocardiogram data including multiple channels; preprocessing the magnetocardiogram data; extracting characteristic indexes from the preprocessed magnetocardiogram data, the characteristic indexes including R-wave amplitude, S-wave amplitude, T-wave amplitude, T-wave direction, ratio of R-wave amplitude to T-wave amplitude, ratio of R-wave amplitude to S-wave amplitude, sum of R-wave amplitude and S-wave amplitude, sum of S-wave amplitude and T-wave amplitude, and consistency index of T-wave direction and main wave direction; constructing and training a machine learning model based on the extracted characteristic indexes for predicting the vascular site of coronary artery stenosis; and using the machine learning model to analyze new magnetocardiogram data to determine whether there is coronary artery disease and the stenotic vascular site.

[0007] As an embodiment of the present invention, in the above method, constructing and training a machine learning model based on the extracted characteristic indexes includes: determining subgroups of four epicardial coronary artery stenosis conditions according to coronary angiography images, including the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk; constructing corresponding machine learning models for each subgroup, and each subgroup is further divided into a training group, an internal validation group, and an external validation group; using the extracted characteristic indexes to predict the incidence of obstruction of each of the four coronary arteries in the training group through a random forest algorithm to determine the stenosis diagnosis models of the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk.

[0008] As an embodiment of the present invention, in the above method, constructing and training a machine learning model based on the extracted characteristic indexes includes: acquiring coronary artery anatomical information, the anatomical information including the diameters, lengths, and numbers of branches of the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk; adding the anatomical information as a feature to the characteristic indexes of the corresponding subgroup; and respectively training the stenosis diagnosis models of the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk based on the extended characteristic indexes, wherein the optimal parameters of the random forest algorithm are determined using a grid search or Bayesian optimization hyperparameter optimization method.

[0009] As an embodiment of the present invention, in the above method, using the machine learning model to analyze new magnetocardiogram data to determine whether there is coronary artery disease and the stenotic vascular site includes: analyzing the new magnetocardiogram data using different stenosis diagnosis models to obtain predicted probability values; based on the coronary artery anatomical information, performing weighted adjustment on the predicted probability values of different stenosis diagnosis models, wherein the larger the vascular diameter, the longer the length, and the more the number of branches, the higher the corresponding weight; and integrating the predicted probability values of the stenosis diagnosis models of each blood vessel after weighted adjustment to obtain the final diagnosis result of coronary artery disease and the stenotic vascular site.

[0010] As an embodiment of the present invention, the integration of the predicted probability values of the stenosis diagnosis models of each blood vessel after weighted adjustment in the above method includes: evaluating the quality of the new magnetocardiogram data, and classifying the data quality into three levels: high, medium, and low according to the signal-to-noise ratio and baseline drift index; dynamically selecting a preset integration strategy according to the quality level of the new magnetocardiogram data, where: if the data quality is high, the predicted probability values of the stenosis diagnosis models of the four blood vessels are weighted and averaged according to the blood vessel diameter, length, and number of branches; if the data quality is medium, the predicted probability values of the stenosis diagnosis models of the four blood vessels are weighted and averaged according to the preset empirical weights; if the data quality is low, the predicted probability values of the stenosis diagnosis models of the four blood vessels are simply averaged.

[0011] As an embodiment of the present invention, after analyzing the new magnetocardiogram data using different stenosis diagnosis models to obtain predicted probability values, the above method further includes: using the isotonic regression method to calibrate the predicted probability of each stenosis diagnosis model.

[0012] According to the second aspect of the present invention, there is provided a data analysis device based on magnetocardiogram, the device includes: a data acquisition unit for acquiring magnetocardiogram data including multiple channels; a preprocessing unit for preprocessing the magnetocardiogram data; a feature index extraction unit for extracting feature indexes from the preprocessed magnetocardiogram data, the feature indexes including R-wave amplitude, S-wave amplitude, T-wave amplitude, T-wave direction, ratio of R-wave amplitude to T-wave amplitude, ratio of R-wave amplitude to S-wave amplitude, sum of R-wave amplitude and S-wave amplitude, sum of S-wave amplitude and T-wave amplitude, and T-wave direction and main wave direction consistency index; a model construction unit for constructing and training a machine learning model based on the extracted feature indexes for predicting the vascular site of coronary artery stenosis; a data analysis unit for analyzing new magnetocardiogram data using the machine learning model to determine whether there is coronary artery disease and the stenotic vascular site.

[0013] Preferably, the above model construction unit includes: a grouping module for determining subgroups of four epicardial coronary artery stenosis conditions according to coronary angiography images, including the left anterior descending branch, left circumflex branch, right coronary artery, and left main trunk; a model construction module for constructing corresponding machine learning models for each subgroup, where each subgroup is further divided into a training group, an internal validation group, and an external validation group; a model training module for predicting the incidence of obstruction of each of the four coronary arteries in the training group using the extracted feature indexes through a random forest algorithm to determine the stenosis diagnosis models of the left anterior descending branch, left circumflex branch, right coronary artery, and left main trunk.

[0014] Preferably, the above-mentioned model construction unit includes: an information acquisition module for acquiring coronary artery anatomical information, where the anatomical information includes the diameters, lengths, and the number of branches of the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk; an index addition module for adding the anatomical information as features to the feature indices of the corresponding subgroups; and a diagnostic model training module for training the stenosis diagnostic models of the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk respectively based on the extended feature indices using the random forest algorithm, where the optimal parameters of the random forest algorithm are determined using grid search or Bayesian optimization hyperparameter optimization methods.

[0015] Preferably, the above-mentioned data analysis unit includes: a prediction module for analyzing new magnetocardiogram data using different stenosis diagnostic models to obtain prediction probability values; an adjustment module for weighted adjustment of the prediction probability values of different stenosis diagnostic models based on coronary artery anatomical information, where the larger the vessel diameter, the longer the length, and the more the number of branches, the higher the corresponding weight; and an integration module for integrating the prediction probability values of the stenosis diagnostic models of each vessel after weighted adjustment to obtain the final diagnosis results of coronary artery disease and the stenosis vessel sites.

[0016] Preferably, the above-mentioned integration module includes: a quality assessment sub-module for quality assessment of new magnetocardiogram data, classifying the data quality into three levels: high, medium, and low according to the signal-to-noise ratio and baseline drift indices; a dynamic selection sub-module for dynamically selecting a preset integration strategy according to the quality level of the new magnetocardiogram data, where: if the data quality is high, the prediction probability values of the stenosis diagnostic models of the four vessels are weighted and averaged according to the vessel diameter, length, and the number of branches; if the data quality is medium, the prediction probability values of the stenosis diagnostic models of the four vessels are weighted and averaged according to preset empirical weights; if the data quality is low, the prediction probability values of the stenosis diagnostic models of the four vessels are simply averaged.

[0017] Preferably, the above-mentioned device further includes: a calibration unit for calibrating the prediction probability of each stenosis diagnostic model using the isotonic regression method.

[0018] According to the third aspect of the present invention, there is provided an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the steps of the above method are implemented.

[0019] According to the fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0020] According to a fifth aspect of the present invention, there is provided a computer program product comprising computer programs / instructions which, when executed by a processor, implement the steps of the above method.

[0021] As can be seen from the above technical solutions, the data analysis method and apparatus based on magnetocardiogram provided by the present invention can more accurately predict the stenotic vascular sites of the coronary arteries by extracting multiple characteristic indexes and combining with a machine learning model, thereby improving the accuracy of the diagnosis of coronary artery diseases. MCG is a highly sensitive detection technology. The present invention uses MCG data for analysis, can detect signs of early coronary artery diseases, and thus achieve early diagnosis and intervention. Therefore, for patients with mild stenosis, coronary angiography or coronary CT is not required, and the toxic side effects of contrast agents and invasive surgeries can be prevented. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings:

[0023] Figure 1 is a schematic flowchart of a data analysis method based on magnetocardiogram provided by an embodiment of the present application;

[0024] Figure 2 is a schematic flowchart of constructing and training a stenosis diagnosis model provided by an embodiment of the present application;

[0025] Figure 3 is a schematic flowchart of constructing and training a stenosis diagnosis model provided by another embodiment of the present application;

[0026] Figure 4 is a schematic flowchart of analyzing new magnetocardiogram data using a machine learning model provided by an embodiment of the present application;

[0027] Figure 5 is a schematic flowchart of integrating the predicted probability values of the stenosis diagnosis models of each blood vessel after weighted adjustment provided by an embodiment of the present application;

[0028] Figure 6 is a schematic structural diagram of a data analysis apparatus based on magnetocardiogram provided by an embodiment of the present application;

[0029] Figure 7 is a schematic structural diagram of a model construction unit provided by an embodiment of the present application;

[0030] Figure 8 It is a schematic structural diagram of a model construction unit provided by another embodiment of the present application;

[0031] Figure 9 It is a schematic structural diagram of a data analysis unit provided by an embodiment of the present application;

[0032] Figure 10 It is a schematic structural diagram of an integration module provided by an embodiment of the present application;

[0033] Figure 11 It is a schematic block diagram of the system composition of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer and more understandable, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Herein, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but not to limit the present invention.

[0035] As Figure 1 shown is a schematic flow diagram of a method for data analysis based on magnetocardiogram provided by an embodiment of the present application. The execution subject of this method can be a client installed in a computer or a server. The method includes the following steps:

[0036] Step S101: Obtain magnetocardiogram data including multiple channels.

[0037] In this step, a multi-channel SERF MCG device can be used to collect cardiac magnetic field data. In this embodiment, data of 36 channels can be extracted, and each channel represents the cardiac magnetic field signal recorded from different spatial positions. The obtained data is one-dimensional MCG data containing time series information. For subsequent training of a machine learning model, the order of magnitude of the magnetocardiogram data obtained in this step is relatively large. Preferably, in this embodiment, 3578 MCG data sets can be used, which include magnetocardiogram data of 1509 normal people and 2069 CAD patients.

[0038] Step S102: Preprocess the magnetocardiogram data.

[0039] The goal of this step of preprocessing is to improve the quality of MCG data and prepare for subsequent feature extraction and model training. The preprocessing here can include, but is not limited to, denoising, baseline correction, filtering, etc. Among them, denoising refers to removing environmental noise and device noise in MCG data, such as power frequency interference, power supply noise, etc. Baseline correction is to correct the baseline drift of MCG data to eliminate the influence of baseline fluctuations on feature extraction. Filtering is to use a filter to remove unwanted frequency components in MCG data, such as high-frequency noise and low-frequency drift. Specifically, for example, a digital low-pass filter with a cut-off frequency of 50 Hz can be used to filter the original MCG data to remove high-frequency noise above 50 Hz; and, a band-pass filter with a frequency range of 1 - 45 Hz can be used to further remove low-frequency drift below 1 Hz and high-frequency noise above 45 Hz in the MCG signal; and, a 50 Hz notch filter can be used to remove power frequency interference; and, the ICA algorithm can be used to remove artificial traces or noise mixed in the MCG bandwidth. ICA is a blind source separation algorithm that can decompose a mixed signal into multiple independent source signals, thereby identifying and removing noise components.

[0040] Preferably, after completing the preprocessing, this embodiment can also evaluate the data quality, such as evaluating indicators such as signal-to-noise ratio, baseline drift, etc.

[0041] Step S103: Extract feature indicators from the preprocessed magnetocardiogram data. The feature indicators include R-wave amplitude (R_amp), S-wave amplitude (S_amp), T-wave amplitude (T_amp), T-wave direction (T_sign), the ratio of R-wave amplitude to T-wave amplitude (ratio_RT), the ratio of R-wave amplitude to S-wave amplitude (ratio_RS), the sum of R-wave amplitude and S-wave amplitude (sum_RS), the sum of S-wave amplitude and T-wave amplitude (sum_ST), and the index of the consistency between T-wave direction and the main wave direction (ZT_sign).

[0042] In this embodiment, the inventor draws on existing electrocardiogram diagnosis experience and combines the characteristics of MCG signals to find indicators that can distinguish normal people from CAD patients from multiple dimensions. First, when selecting feature indicators in this embodiment, the established nomenclature of electrocardiogram and ischemic indicators are referred to, and combined with the results of coronary angiography (CAG) to ensure that the selected feature indicators are related to the pathophysiological mechanism of coronary artery disease.

[0043] Specifically, single-index analysis is first performed. By comparing the index values of normal individuals and CAD patients on the same channel, it is found that there are significant differences in amplitude, direction, and ratio among these characteristic indices. For example, the T-wave amplitude (T_amp) on the 30th channel of LAD patients shows depression, while that of normal individuals does not, which is similar to the phenomenon of T-wave depression in electrocardiograms. Then, multi-index analysis is carried out. By comprehensively analyzing multiple indices, it is found that there are also significant differences in the combinations of these indices between normal individuals and CAD patients. This indicates that the combined use of multiple indices can improve the accuracy of diagnosis. Finally, multi-channel analysis is performed. The distribution patterns of these indices on 36 channels are analyzed, and it is found that there are significant differences in the distribution patterns between CAD patients and normal individuals. For example, the channels with the highest R-wave amplitude (R_amp) in normal individuals are concentrated in the 20th - 22nd and 26th - 29th channels, while the channels with the largest R-wave amplitude in LAD patients are dispersed to the 30th channel, which indicates that the use of multi-channel information can better capture the abnormal changes in cardiac electrophysiological activities.

[0044] In summary, through various aspects of analysis and comparison, the inventor has proven that these 9 indices can effectively distinguish between normal individuals and CAD patients and have potential clinical application value. The selection of these indices is not based on pure mathematical derivation, but combines knowledge of cardiac electrophysiology, electrocardiogram diagnosis experience, and the characteristics of MCG signals, so it has more clinical significance.

[0045] In this embodiment, these 9 indices are extracted for each channel. Therefore, each data sample will finally have 9 * 36 = 324 characteristic variables. Here, the data sample refers to the magnetocardiogram data of a normal individual or a CAD patient.

[0046] Step S104: Based on the extracted characteristic indices, construct and train a machine learning model for predicting the vascular site of coronary artery stenosis. This step realizes a reliable mapping from MCG signals to the diagnosis of coronary artery stenosis through the deep combination of machine learning and clinical knowledge.

[0047] Preferably, as Figure 2 shown, this step can further include the following sub-steps:

[0048] Step S201: Determine subgroups of four epicardial coronary artery stenosis conditions according to coronary angiography images, including the left anterior descending branch, left circumflex branch, right coronary artery, and left main trunk.

[0049] In this embodiment, first, patients are grouped according to coronary angiography images, and this grouping reflects the stenosis conditions of epicardial coronary arteries. The four subgroups are: left anterior descending artery (LAD), left circumflex artery (LCX), right coronary artery (RCA), and left main coronary artery (LM). The stenosis conditions of epicardial coronary arteries can adopt the internationally common CAG diagnostic criteria, that is, a lumen stenosis ≥ 50% is defined as significant stenosis. In addition, patients with multi-vessel lesions need to be excluded to ensure the specificity of the subgroups.

[0050] Preferably, while grouping, a mapping relationship can be established between the stenosis positions determined by CAG and the spatial distribution of MCG channels, and it is ensured that the time interval between the MCG signal acquisition period and the CAG examination time ≤ 72 hours.

[0051] This step provides a training data basis that conforms to the electro-anatomical laws of the heart for the subsequent machine learning model through strict imaging criteria and physiological mapping.

[0052] Step S202: Construct corresponding machine learning models for each subgroup, and each subgroup is further divided into a training group, an internal validation group, and an external validation group.

[0053] Since different coronary arteries supply different myocardial regions (for example, the LAD supplies the anterior wall and the RCA supplies the inferior wall), the abnormal patterns of MCG signals have spatial specificity. For example, in patients with LAD stenosis, significant T-wave depression (amplitude decrease > 40%) appears in channels 20 - 30; in patients with RCA stenosis, T-wave inversion appears in channels 5 - 15. It is difficult for a single model to simultaneously capture the unique signal patterns of different vascular lesions, while sub-models can specifically learn the differences in feature weights and spatial distributions. In addition, by constructing corresponding machine learning models for each subgroup respectively, each model can focus on a single vascular lesion, avoiding prediction bias caused by sample size differences.

[0054] Step S203: Through the random forest algorithm, use the extracted feature indicators to predict the incidence of obstruction of each of the four coronary arteries in the training group, so as to determine the stenosis diagnosis models for the left anterior descending artery, left circumflex artery, right coronary artery, and left main coronary artery.

[0055] This step deeply combines the non-linear modeling ability of the random forest with the spatial specificity of MCG signals to achieve accurate localization diagnosis of coronary artery stenosis. Among them, in the training set, in the LAD, LCX, RCA, and LM cohorts, the accuracies of diagnosing stenosis in specific vascular regions are 0.987, 0.979, 0.985, and 0.880 respectively. In the internal validation set, they are 0.928, 0.940, 0.922, and 0.880 respectively. In the external validation, the accuracies are 0.930, 0.881, 0.908, and 0.939 respectively.

[0056] Preferably, as Figure 3 shown, this step S104 may further include the following steps:

[0057] Step S301: Obtain coronary artery anatomical information, where the anatomical information includes the diameters, lengths, and numbers of branches of the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk.

[0058] Here, the coronary artery anatomical information can be obtained through coronary angiography or CT coronary angiography. After obtaining this anatomical information, continuous variables such as diameters and lengths can be normalized to the [0,1] interval to avoid the influence of feature scale differences on model training, while the number of branches, as a discrete variable, is directly input into the model.

[0059] Step S302: Add the anatomical information as features to the feature indicators of the corresponding subgroups.

[0060] In this step, the anatomical information is added to the feature indicators of the corresponding subgroups in the following manner:

[0061] LAD subgroup: MCG signal features + LAD anatomical features;

[0062] LCX subgroup: MCG signal features + LCX anatomical features;

[0063] RCA subgroup: MCG signal features + RCA anatomical features;

[0064] LM subgroup: MCG signal features + LM anatomical features.

[0065] Step S303: Based on the extended feature indicators, use the random forest algorithm to train the stenosis diagnosis models for the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk respectively, where the optimal parameters of the random forest algorithm are determined using grid search or Bayesian optimization hyperparameter optimization methods.

[0066] In this embodiment, by combining the anatomical information with the MCG signal features, the deficiency of a single data source is made up, and the diagnostic accuracy of the model is significantly improved. Grid search and Bayesian optimization ensure the optimal parameter combination of the random forest algorithm, avoiding overfitting or underfitting. In addition, modeling separately for each coronary artery and clarifying the diagnostic basis for each blood vessel can facilitate the understanding and application by clinicians. This multi-data-source fusion analysis method provides more comprehensive technical support for the non-invasive diagnosis of coronary artery diseases.

[0067] It should be noted that Figure 3 the corresponding embodiment is related to Figure 2Parallel, which adds anatomical information as a feature and introduces a hyperparameter optimization method, making the features of the model more comprehensive and the diagnostic ability stronger.

[0068] Step S105: Analyze the new magnetocardiogram data using the machine learning model to determine whether there is coronary artery disease and the location of the stenotic blood vessels.

[0069] Preferably, when the feature index in the foregoing steps includes anatomical information, as Figure 4 shown, this step may further include the following sub-steps:

[0070] Step S1051: Analyze the new magnetocardiogram data using different stenosis diagnosis models to obtain predicted probability values.

[0071] Each coronary artery (left anterior descending artery LAD, left circumflex artery LCX, right coronary artery RCA, left main coronary artery LM) corresponds to an independent machine learning model (stenosis diagnosis model). Input the new magnetocardiogram data and the extended feature index (including anatomical information) into the diagnosis model, and then each stenosis diagnosis model outputs the stenosis probability value of the corresponding blood vessel.

[0072] Preferably, after analyzing the new magnetocardiogram data using different stenosis diagnosis models to obtain predicted probability values in this step, it may further include: using the isotonic regression method to calibrate the predicted probabilities of each stenosis diagnosis model.

[0073] The predicted probability values output by the machine learning model usually represent the likelihood of an event occurring. However, these probability values may not exactly conform to the true probability distribution. Through the probability calibration method, the predicted probabilities output by the model can be made closer to the true probabilities. For example, when the model predicts a probability of 80%, the proportion of the actual events occurring should also be close to 80%. Isotonic Regression is a non-parametric calibration method that assumes the relationship between the predicted probability and the true probability is monotonically increasing. It calibrates the probabilities output by the model by fitting a monotonic function. It does not require strong assumptions about the data distribution and can effectively handle the bias of the probabilities output by the model, especially suitable for cases with a large sample size.

[0074] Step S1052: Based on the coronary artery anatomical information, weight-adjust the predicted probability values of different stenosis diagnosis models, where the larger the blood vessel diameter, the longer the length, and the more the number of branches, the higher the corresponding weight.

[0075] In this step, the basis for weighted adjustment is coronary artery anatomical information (diameter, length, number of branches), which reflects the importance of the blood vessel and the blood supply range: The larger the diameter: the stronger the blood supply capacity of the blood vessel, and the greater the impact of stenosis on the myocardium. The longer the length: the wider the coverage area of the blood vessel, and the higher the clinical significance of stenosis. The more the number of branches: the more complex the blood supply area of the blood vessel, and the more significant the impact of stenosis.

[0076] Assume that the anatomical weight of each blood vessel is w i , and its calculation formula is:

[0077] w i = α × diameter + β × length + γ × number of branches;

[0078] Among them, α, β, and γ are weight coefficients (which can be adjusted through clinical data).

[0079] The predicted probability value after weighting is:

[0080] P i ’ = P i × w i ;

[0081] Among them, P i is the original predicted probability, and P i ’ is the probability after weighting.

[0082] The following uses a specific example to further describe this step:

[0083] Assume the anatomical information is as follows:

[0084] LAD: diameter = 3.5mm, length = 12cm, number of branches = 4;

[0085] LCX: diameter = 2.5mm, length = 8cm, number of branches = 2;

[0086] RCA: diameter = 3.0mm, length = 10cm, number of branches = 3;

[0087] LM: diameter = 4.5mm, length = 1.2cm, number of branches = 2;

[0088] Weight calculation:

[0089] w LAD = 0.5 × 3.5 + 0.3 × 12 + 0.2 × 4 = 5.85;

[0090] w LCX = 0.5 × 2.5 + 0.3 × 8 + 0.2 × 2 = 4.15;

[0091] w RCA= 0.5 × 3.0 + 0.3 × 10 + 0.2 × 3 = 5.3;

[0092] w LM = 0.5 × 4.5 + 0.3 × 1.2 + 0.2 × 2 = 3.24;

[0093] Weighted adjusted probability:

[0094] P LAD = 0.85 × 5.85 = 4.9725;

[0095] P LCX = 0.45 × 4.15 = 1.8675;

[0096] P RCA = 0.65 × 5.3 = 3.445;

[0097] P LM = 0.20 × 3.24 = 0.648.

[0098] Step S1053: Integrate the predicted probability values of each vascular model after weighted adjustment to obtain the final diagnosis result of coronary artery disease and its stenotic vascular sites.

[0099] In this step, the overall stenosis probability can be calculated based on the weighted probability values, and then the vessel with the highest weighted probability value can be selected as the main diagnostic basis. Through weighted adjustment, the importance of coronary artery anatomical features in stenosis diagnosis is fully considered, enhancing the clinical relevance of the diagnosis. By using the prediction results of different vascular models, the limitations of a single model can be avoided, improving the comprehensiveness and accuracy of the diagnosis. Additionally, the present application can dynamically adjust the weights according to the anatomical features of different patients to meet the needs of personalized diagnosis.

[0100] Further preferably, as Figure 5 shown, the integration of the predicted probability values of the stenosis diagnosis models of each vessel after weighted adjustment in the above step S1053 can further include:

[0101] Step S10531: Evaluate the quality of the new magnetocardiogram data, and classify the data quality into three levels: high, medium, and low according to the signal-to-noise ratio and baseline drift indicators.

[0102] In this embodiment, an evaluation mechanism for MCG data quality is further introduced, and the integration strategy is dynamically adjusted according to the data quality to improve the reliability and robustness of the diagnosis result. The signal-to-noise ratio (SNR) is the ratio of the effective component of the signal to the noise component. The higher the SNR, the better the data quality. Baseline drift can evaluate the stability of the signal baseline, and the smaller the drift, the higher the data quality.

[0103] According to the SNR and baseline drift, the data quality is divided into three levels: high, medium, and low. For example:

[0104] High quality: SNR > 20 dB and baseline drift < 0.1 μV;

[0105] Medium quality: 10 dB ≤ SNR ≤ 20 dB or baseline drift is between 0.1 - 0.3 μV;

[0106] Low quality: SNR < 10 dB or baseline drift > 0.3 μV.

[0107] Of course, the above classification is only an example, and those skilled in the art can adjust the basis for level division according to needs.

[0108] Step S10532: Dynamically select a preset integration strategy according to the quality level of the new magnetocardiogram data.

[0109] Step S10533: If the data quality is high, perform a weighted average on the predicted probability values of the stenosis diagnosis models of the four blood vessels according to the blood vessel diameter, length, and number of branches.

[0110] Step S10534: If the data quality is medium, perform a weighted average on the predicted probability values of the stenosis diagnosis models of the four blood vessels according to the preset empirical weights.

[0111] Step S10535: If the data quality is low, perform a simple average on the predicted probability values of the stenosis diagnosis models of the four blood vessels.

[0112] As can be seen from the above, in this embodiment, the integration strategy is dynamically adjusted according to the data quality, which can ensure the reliability and robustness of the diagnosis result. Under high-quality data conditions, weighted averaging is performed in combination with anatomical information to improve the diagnostic accuracy. Under low-quality data conditions, a simple average strategy is adopted to avoid error amplification caused by improper weight allocation. Through sub-steps S1053 - S10535, the integration process of step S1053 is further optimized. This enables this embodiment to dynamically select the integration strategy according to the magnetocardiogram data quality, making full use of the advantages of high-quality data and ensuring the stability of the diagnosis result under low-quality data conditions, providing stronger technical support for the non-invasive diagnosis of coronary artery disease.

[0113] As described above, the data analysis method based on magnetocardiogram provided by the present invention can more accurately predict the vascular site of coronary artery stenosis by extracting multiple characteristic indexes and combining with a machine learning model, thereby improving the accuracy of the diagnosis of coronary artery disease. MCG is a highly sensitive detection technology. By analyzing MCG data, the present invention can detect the signs of early coronary artery disease, thereby realizing early diagnosis and intervention. Therefore, for patients with mild stenosis, coronary angiography or coronary CT is not required, and the toxic side effects of contrast agents and invasive surgeries can be prevented.

[0114] As Figure 6 shown in the structural schematic diagram of a data analysis device based on magnetocardiogram provided by an embodiment of the present application, the device includes: a data acquisition unit 610, a preprocessing unit 620, a characteristic index extraction unit 630, a model construction unit 640, and a data analysis unit 650, which are connected in sequence. Among them:

[0115] The data acquisition unit 610 is configured to acquire magnetocardiogram data including multiple channels.

[0116] The preprocessing unit 620 is configured to preprocess the magnetocardiogram data.

[0117] The characteristic index extraction unit 630 is configured to extract characteristic indexes from the preprocessed magnetocardiogram data, and the characteristic indexes include R wave amplitude, S wave amplitude, T wave amplitude, T wave direction, ratio of R wave amplitude to T wave amplitude, ratio of R wave amplitude to S wave amplitude, sum of R wave amplitude and S wave amplitude, sum of S wave amplitude and T wave amplitude, and consistency index of T wave direction and main wave direction.

[0118] The model construction unit 640 is configured to construct and train a machine learning model based on the extracted characteristic indexes for predicting the vascular site of coronary artery stenosis.

[0119] The data analysis unit 650 is configured to analyze new magnetocardiogram data by using the machine learning model to determine whether there is coronary artery disease and the stenotic vascular site.

[0120] Preferably, as Figure 7 shown, the above model construction unit 640 includes:

[0121] The grouping module 641 is configured to determine subgroups of four epicardial coronary artery stenosis conditions according to coronary angiography images, including the left anterior descending branch, the left circumflex branch, the right coronary artery, and the left main trunk.

[0122] The model construction module 642 is configured to construct corresponding machine learning models for each subgroup, and each subgroup is further divided into a training group, an internal validation group, and an external validation group.

[0123] A model training module 643, configured to use the random forest algorithm to predict the incidence of blockage of each of the four coronary arteries in the training group by using the extracted feature metrics, so as to determine the stenosis diagnosis models for the left anterior descending artery, the left circumflex artery, the right coronary artery, and the left main trunk.

[0124] Preferably, as Figure 8 shown, the above-mentioned model construction unit 640 includes:

[0125] An information acquisition module 644, configured to acquire coronary artery anatomical information, where the anatomical information includes the diameters, lengths, and numbers of branches of the left anterior descending artery, the left circumflex artery, the right coronary artery, and the left main trunk.

[0126] An index addition module 645, configured to add the anatomical information as a feature to the feature metrics of the corresponding subgroup.

[0127] A diagnosis model training module 646, configured to respectively train the stenosis diagnosis models for the left anterior descending artery, the left circumflex artery, the right coronary artery, and the left main trunk based on the extended feature metrics by using the random forest algorithm, where the optimal parameters of the random forest algorithm are determined by using a grid search or a Bayesian optimization hyperparameter optimization method.

[0128] Preferably, as Figure 9 shown, the above-mentioned data analysis unit 650 includes:

[0129] A prediction module 651, configured to analyze new magnetocardiogram data by using different stenosis diagnosis models to obtain prediction probability values.

[0130] An adjustment module 652, configured to perform weighted adjustment on the prediction probability values of different stenosis diagnosis models based on coronary artery anatomical information, where the larger the blood vessel diameter, the longer the length, and the more the number of branches, the higher the corresponding weight.

[0131] An integration module 653, configured to integrate the prediction probability values of the stenosis diagnosis models of each blood vessel after weighted adjustment to obtain the final diagnosis result of coronary artery disease and the stenosis blood vessel site.

[0132] Preferably, as Figure 10 shown, the above-mentioned integration module 653 may further include:

[0133] A quality assessment sub-module 6531, configured to perform quality assessment on new magnetocardiogram data, and classify the data quality into three levels: high, medium, and low according to the signal-to-noise ratio and baseline drift indicators.

[0134] The dynamic selection sub-module 6532 is used to dynamically select a preset integration strategy according to the quality level of the new magnetocardiogram data, where: if the data quality is high, the predicted probability values of the stenosis diagnosis models of the four blood vessels are weighted and averaged according to the blood vessel diameter, length, and number of branches; if the data quality is medium, the predicted probability values of the stenosis diagnosis models of the four blood vessels are weighted and averaged according to the preset empirical weights; if the data quality is low, the predicted probability values of the stenosis diagnosis models of the four blood vessels are simply averaged.

[0135] Preferably, the above device further includes: a calibration unit for calibrating the predicted probability of each stenosis diagnosis model using the isotonic regression method.

[0136] As can be seen from the above technical solutions, the data analysis device based on magnetocardiogram provided by the present invention can more accurately predict the vascular sites of coronary artery stenosis by extracting multiple characteristic indexes and combining machine learning models, thereby improving the accuracy of coronary artery disease diagnosis. MCG is a highly sensitive detection technology. The present invention uses MCG data for analysis, can detect the signs of early coronary artery disease, and thus achieve early diagnosis and intervention. Therefore, for patients with mild stenosis, coronary angiography or coronary CT is not required, and the toxic side effects of contrast agents and invasive surgeries can be prevented.

[0137] An embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above method is implemented.

[0138] An embodiment of the present invention further provides a computer program product, including computer programs / instructions. When the computer programs / instructions are executed by a processor, the steps of the above method are implemented.

[0139] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program for executing the above method.

[0140] As Figure 11 shown, the electronic device 600 may further include: a communication module 110, an input unit 120, an audio processor 130, a display 160, and a power supply 170. It should be noted that the electronic device 600 does not necessarily have to include all the components shown in Figure 11 ; in addition, the electronic device 600 may further include components not shown in Figure 11 , and reference may be made to the prior art.

[0141] As Figure 11As shown, the central processing unit 100, sometimes also referred to as a controller or operation control, may include a microprocessor or other processor device and / or logic device. The central processing unit 100 receives inputs and controls the operation of various components of the electronic device 600.

[0142] Among them, the memory 140 can be, for example, one or more of a buffer, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. It can store the above information related to failures, and can also store programs for executing relevant information. And the central processing unit 100 can execute the programs stored in the memory 140 to implement information storage or processing, etc.

[0143] The input unit 120 provides inputs to the central processing unit 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to supply power to the electronic device 600. The display 160 is used to display display objects such as images and texts. The display can be, for example, an LCD display, but is not limited thereto.

[0144] The memory 140 can be a solid-state memory. For example, it can be a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that stores information even when powered off, can be selectively erased, and has more data. Examples of such a memory are sometimes referred to as EPROMs, etc. The memory 140 can also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 can include an application / function storage unit 142, which is used to store application programs and function programs or the processes for operating the electronic device 600 through the central processing unit 100.

[0145] The memory 140 can also include a data storage unit 143, which is used to store data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 144 of the memory 140 can include various drivers of the electronic device for communication functions and / or for performing other functions of the electronic device (such as a messaging application, an address book application, etc.).

[0146] The communication module 110 is a transmitter / receiver that transmits and receives signals via the antenna 111. The communication module 110 (transmitter / receiver) is coupled to the central processing unit 100 to provide input signals and receive output signals, which can be the same as in the case of a conventional mobile communication terminal.

[0147] Based on different communication technologies, in the same electronic device, multiple communication modules 110 can be provided, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. The communication module 110 (transmitter / receiver) is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide an audio output via the speaker 131 and receive an audio input from the microphone 132, so as to implement normal telecommunication functions. The audio processor 130 can include any suitable buffers, decoders, amplifiers, etc. Additionally, the audio processor 130 is also coupled to a central processor 100, enabling recording on the local device through the microphone 132 and playing back the sounds stored on the local device through the speaker 131.

[0148] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0150] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the steps of the function specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 in one block or a plurality of blocks.

[0152] In the present invention, specific embodiments are used to illustrate the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A data analysis method based on magnetocardiography, characterized in that: The method comprises: Acquiring magnetocardiographic data containing multiple channels; Preprocessing the magnetocardiogram data; Extracting characteristic indicators from the pre-processed magnetocardiogram data, the characteristic indicators include R wave amplitude, S wave amplitude, T wave amplitude, T wave direction, the ratio of R wave amplitude to T wave amplitude, the ratio of R wave amplitude to S wave amplitude, the sum of R wave amplitude and S wave amplitude, the sum of S wave amplitude and T wave amplitude, and the consistency index of T wave direction and main wave direction; Based on the extracted characteristic indicators, a machine learning model is constructed and trained to predict the vascular location of coronary artery stenosis; The machine learning model is used to analyze new magnetocardiographic data to determine whether coronary artery disease and the location of stenosis are present.

2. The magnetocardiogram-based data analysis method according to claim 1, characterized in that: The step of constructing and training a machine learning model based on the extracted feature indicators includes: Four subgroups of epicardial coronary artery stenosis were determined based on coronary angiographic images, including the left anterior descending artery, left circumflex artery, right coronary artery, and left main artery; A corresponding machine learning model was constructed for each subgroup, where each subgroup was divided into a training group, an internal validation group, and an external validation group; The extracted feature indices were used to predict the incidence of coronary artery occlusion in each of the four coronary arteries in the training group using the random forest algorithm to determine the stenosis diagnostic model for the left anterior descending artery, left circumflex artery, right coronary artery, and left main trunk.

3. The data analysis method based on magnetocardiography according to claim 1, characterized in that: The step of constructing and training a machine learning model based on the extracted feature indicators includes: Obtaining vascular stenosis information of major branches of the coronary arteries, including the diameter, length, and number of branches of the left anterior descending artery, the left circumflex artery, the right coronary artery, and the left main trunk; The coronary angiography information was added as a label to the characteristic indicators of the corresponding subgroups for supervised machine learning; Based on the expanded feature indicators, the random forest algorithm was used to train the stenosis diagnosis models of the left anterior descending artery, left circumflex artery, right coronary artery and left main trunk, respectively. The optimal parameters of the random forest algorithm were determined using grid search or Bayesian optimization hyperparameter optimization methods.

4. The magnetocardiogram-based data analysis method according to claim 3, characterized in that: The use of the machine learning model to analyze new magnetocardiogram data to determine whether there is coronary artery disease and the location of stenosis includes: The new magnetocardiographic data were analyzed using different stenosis diagnostic models to obtain predicted probability values; Based on the anatomical information of the coronary arteries, the predicted probability values ​​of different stenosis diagnostic models are weighted and adjusted. The larger the vessel diameter, the longer the length, and the more branches, the higher the corresponding weight. The predicted probability values ​​of the weighted and adjusted stenosis diagnostic models of each blood vessel are integrated to obtain the final diagnosis results of coronary artery disease and stenosis blood vessel locations.

5. The magnetocardiogram-based data analysis method according to claim 4, characterized in that: The step of integrating the predicted probability values ​​of the stenosis diagnosis models of the various blood vessels after weighted adjustment includes: The quality of new magnetocardiographic data was assessed and the data quality was classified into three levels: high, medium, and low according to the signal-to-noise ratio and baseline drift indicators; A preset integration strategy is dynamically selected based on the quality level of the new magnetocardiographic data, where: If the data quality is high, the predicted probability values ​​of the stenosis diagnosis model of the four vessels are weighted averaged according to the vessel diameter, length, and number of branches; If the data quality is medium, the predicted probability values ​​of the stenosis diagnosis model of the four vessels are weighted averaged according to the pre-set empirical weights; If the data quality is low, the predicted probability values ​​of the stenosis diagnosis models of the four vessels are simply averaged.

6. The magnetocardiogram-based data analysis method according to claim 4, characterized in that: After analyzing the new magnetocardiographic data using different stenosis diagnostic models to obtain the predicted probability value, the method further includes: using an rank-preserving regression method to calibrate the predicted probability of each stenosis diagnostic model.

7. A data analysis device based on magnetocardiography, characterized in that: The device comprises: A data acquisition unit, used for acquiring magnetocardiographic data including a plurality of channels; A preprocessing unit, used for preprocessing the magnetocardiogram data; a characteristic index extraction unit, used for extracting characteristic indexes from the preprocessed magnetocardiogram data, wherein the characteristic indexes include R wave amplitude, S wave amplitude, T wave amplitude, T wave direction, the ratio of R wave amplitude to T wave amplitude, the ratio of R wave amplitude to S wave amplitude, the sum of R wave amplitude and S wave amplitude, the sum of S wave amplitude and T wave amplitude, and the consistency index of T wave direction and main wave direction; A model building unit, used to build and train a machine learning model based on the extracted characteristic indicators, so as to predict the vascular location of coronary artery stenosis; A data analysis unit is used to analyze new magnetocardiographic data using the machine learning model to determine whether coronary artery disease and stenotic blood vessel locations exist.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images