Multi-domain feature extraction method and system for multi-channel magnetocardiogram

Abstract

The invention discloses a multi-domain feature extraction method and system for a multi-channel magnetocardiogram, and the method comprises the steps: firstly collecting a multi-channel magnetocardiogram signal through a magnetocardiogram measurement device, checking the data quality, removing invalid channels, and removing the power frequency interference and baseline drift in the magnetocardiogram signal; a waveform positioning algorithm is applied to divide a magnetocardiogram signal into a P wave band, a QRS wave band and a T wave band. Respectively extracting morphological parameters of P, QRS and T wave bands of the magnetocardiogram signals as time domain feature matrixes; decomposing P, QRS and T magnetocardiogram signals by using wavelet packet transformation, and respectively extracting frequency domain characteristics of each wave band; and according to the energy distribution of the space magnetic field, calculating the energy distribution of different frequencies in the space by utilizing the frequency domain characteristics of each channel, and extracting the specific space domain characteristics of the multi-channel magnetocardiogram data. The invention provides a multi-channel time domain, frequency domain and space domain feature extraction method for overcoming the defects of an existing multi-channel magnetocardiogram signal feature extraction method, and the method has important contribution to a modeling test based on magnetocardiogram equipment.

Description

Technical Field

[0001] The present invention belongs to the field of magnetocardiogram signal processing and feature extraction, and in particular relates to a multi-domain feature extraction method and system for a multi-channel magnetocardiogram. Background Art

[0002] Because rats resemble humans in cardiovascular and cerebrovascular systems, drug metabolism, and other aspects, and because they are highly reproducible, rat modeling experiments can be used to establish disease models for arrhythmias and other conditions. These models are widely used in the study of disease mechanisms, diagnosis, treatment, and prevention. Measuring rat magnetocardiography (MCG) helps deepen understanding of the pathological mechanisms of heart disease, providing an important experimental foundation and tool for its clinical application.

[0003] In rat modeling experiments based on magnetocardiography (MCG), the extraction and interpretation of characteristic parameters from MCG datasets can aid in the detection of heart disease and help advance research into disease mechanisms. However, due to the weak rat MCG signal and the limited number of channels, current methods for extracting features from one-dimensional butterfly plots and two-dimensional images of MCGs are difficult to directly apply to the task of extracting features from rat MCGs. These methods cannot fully utilize the temporal and spatial information in rat MCGs, limiting the auxiliary role of MCG-based rat modeling experiments in disease mechanism research. Summary of the Invention

[0004] To address the technical issues of difficulty in extracting magnetocardiographic features from magnetocardiographic experiments and the inability to fully exploit the time, frequency, and spatial domain information of the magnetocardiogram, the present invention provides a multi-domain feature extraction method and system for multi-channel magnetocardiograms. This method reduces noise in the magnetocardiograph and extracts multi-dimensional features of the magnetocardiograph. These features include the time domain waveform features of each band, the frequency domain features of each band based on wavelet packet transform, and the spatial domain features of the energy distribution of each band.

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

[0006] A multi-domain feature extraction method for a multi-channel magnetocardiogram, the method comprising:

[0007] S10: Collect multi-channel cardiac magnetometry signals through a cardiac magnetometry measurement device, check data quality, remove invalid channels, and remove power frequency interference and baseline drift in the cardiac magnetometry signals;

[0008] S20: Apply waveform localization algorithm to divide the magnetic cardiogram signal into P wave, QRS wave and T wave;

[0009] S30: extracting the P-band, QRS-band, and T-band morphological parameters of the magnetic cardiogram signal as time domain feature matrices;

[0010] S40: Decompose the P-band, QRS-band, and T-band magnetocardiographic signals using wavelet packet transform, and extract the frequency domain features of each band respectively;

[0011] S50: Calculate the energy distribution of different frequencies in space using the frequency domain features of each channel, and extract the spatial features of each band.

[0012] Preferably, the S10 includes:

[0013] S11, using an optically pumped magnetometer to collect multi-channel magnetocardiographic signals, and filtering out invalid channels based on the correlation of the autocorrelation functions between the channels; when the average correlation coefficient between the autocorrelation functions of a certain channel and other channels is lower than a set threshold, the channel is considered to be an invalid channel and is removed;

[0014] S12, performing wavelet packet decomposition on the signal, setting the approximate coefficients representing baseline drift in the wavelet packet to zero; identifying the wavelet packet coefficients containing power frequency noise based on the kurtosis and kurtosis of the wavelet packet coefficients in the frequency domain, and setting the coefficients to zero; performing soft threshold denoising on all wavelet packet coefficients, and then reconstructing to obtain the denoised magnetocardiographic signal;

[0015] Wherein, the S11 includes:

[0016]

[0017] Among them, i represents the channel to be judged, j represents other channels except channel i, R i , R j represents the autocorrelation function between channel i and channel j, and L represents the total length of the acquired data;

[0018] The S12 includes:

[0019]

[0020] Among them, sig represents the denoised magnetic cardiogram signal, μ i represents the i-th wavelet packet coefficient after denoising.

[0021] Preferably, the S20 includes:

[0022] S21: Locating the R peak of the magnetocardiographic signal using a peak-finding algorithm;

[0023] S22: After locating the R peak, set the R wave to zero and perform peak location again. Find the P wave position at a sampling point before the Q wave, and find the T wave position at b sampling points after the Q wave, completing the location of the P, QRS, and T wave bands.

[0024] S23: Complete the P, QRS, and T band magnetic cardiogram signals u according to the set waveform width P,i (t),u R，i(t),u T,i (t), where i represents the i-th channel and t represents the t-th sampling point.

[0025] Preferably, the S30 includes:

[0026] Normalize the amplitude of each band to extract the parameters of each band, including kurtosis, skewness, pulse factor, shape factor, peak factor and margin factor;

[0027] The feature parameters are combined into a time domain feature matrix ET of size 3*6 P ,ET R ,ET T , which consists of 6 waveform characteristic parameters of P, QRS, and T bands;

[0028] Among them, normalizing the amplitude of each band includes:

[0029]

[0030] Among them, u P ,u R ,u T represent the extracted P, QRS, and T wave bands, respectively;

[0031] Calculations of kurtosis, skewness, impulse factor, shape factor, crest factor, and margin factor include:

[0032]

[0033] in, represents the average value of the signal s(t), and L represents the total length of the acquired data.

[0034] Preferably, the S40 includes:

[0035] S41: using wavelet packet transform to decompose the P, QRS, and T bands of each cardiac cycle of each channel respectively, and obtain the wavelet packet coefficients of each band;

[0036] S42: Based on the wavelet packet decomposition results, the energy weights of the wavelet packet coefficients in the P, QRS, and T bands are calculated respectively to obtain the energy distribution of each band in different frequency bands; the obtained frequency domain parameters are combined into a matrix to characterize the frequency domain characteristics of the single-channel magnetocardiographic signal;

[0037] Among them, according to the wavelet packet decomposition results, the energy weights of each wavelet packet coefficient in the P, QRS, and T bands are calculated respectively, and the energy distribution of each band in different frequency bands is obtained, including:

[0038]

[0039] Among them, EFP i,j , EF R i,j , EF Ti,j Respectively represent the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, μ Pi,j , μ Ri,j , μ Ti,j represents the jth wavelet packet coefficient of the i-th channel of the P, QRS, and T bands respectively;

[0040] The obtained frequency domain parameters are combined into a matrix to characterize the frequency domain characteristics of the single-channel magnetic cardiogram signal, including:

[0041] The obtained frequency domain parameters EF Pi,j , EF Ri,j , EF Ti,j Form three matrices EF of size N*M respectively P , EF R , EF T , respectively characterizing the frequency domain characteristics of the P band, QRS band and T band of the single-channel magnetic cardiogram signal, where M represents the number of sensors and N represents the number of wavelet packet coefficients obtained by decomposition.

[0042] Preferably, the S50 includes:

[0043] Using multi-channel magnetocardiographic signals and the sensor array arrangement, the energy distribution matrix of each frequency band at different bands is calculated to obtain the energy spatial distribution of each band, that is, the spatial domain characteristics;

[0044] The spatial domain features are composed of three N*M matrices, representing the spatial domain features of P wave, QRS wave, and T wave respectively, where N is the number of sensor channels and M is the number of wavelet packet coefficients;

[0045] Among them, the energy distribution matrix of each frequency band at different bands is calculated to obtain the energy spatial distribution of each band, that is, the spatial domain characteristics, including:

[0046]

[0047] Among them, ES P i,j , ES R i,j , ES Ti,j represents the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, respectively, ||μ Pi,j ||,||μ Ri,j ||,||μ Ti,j || represents the bi-norm of the j-th wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, respectively.

[0048] The present invention also provides a multi-domain feature extraction system for a multi-channel magnetocardiogram, the system being used to implement the aforementioned method, the system comprising: an acquisition module, a division module, an extraction module, a decomposition module, and a calculation module;

[0049] The acquisition module is used to collect multi-channel cardiac magnetometry signals through a cardiac magnetometry measurement device, check data quality and eliminate invalid channels, and remove power frequency interference and baseline drift in the cardiac magnetometry signals;

[0050] The division module is used to apply a waveform positioning algorithm to divide the magnetic cardiogram signal into a P wave band, a QRS wave band, and a T wave band;

[0051] The extraction module is used to extract the P-band, QRS-band, and T-band morphological parameters of the magnetocardiographic signal as time domain feature matrices;

[0052] The decomposition module is used to decompose the P-band, QRS-band, and T-band magnetocardiographic signals using wavelet packet transform to extract frequency domain features of each band;

[0053] The calculation module is used to calculate the energy distribution of different frequencies in space by using the frequency domain characteristics of each channel and extract the spatial characteristics of each band.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention proposes a multi-domain feature extraction method for multi-channel magnetocardiograms. By combining waveform segmentation, the time domain features and frequency domain features of the P, QRS, and T waveforms of the magnetocardiogram are extracted respectively. Based on the spatial magnetic field energy distribution, the unique spatial domain features of multi-channel magnetocardiogram data are extracted. This method has important significance for model experiments based on magnetocardiogram equipment and their data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 The present invention is a flowchart of a multi-domain feature extraction method for a multi-channel magnetocardiogram. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] Example 1

[0061] This embodiment provides a multi-domain feature extraction method for a multi-channel magnetocardiogram, the method comprising:

[0062] S10: Collect multi-channel cardiac magnetometry signals through a cardiac magnetometry measurement device, check data quality, remove invalid channels, and remove power frequency interference and baseline drift in the cardiac magnetometry signals;

[0063] S20: Apply waveform localization algorithm to divide the magnetic cardiogram signal into P wave, QRS wave and T wave;

[0064] S30: extracting the P-band, QRS-band, and T-band morphological parameters of the magnetic cardiogram signal as time domain feature matrices;

[0065] S40: Decompose the P-band, QRS-band, and T-band magnetocardiographic signals using wavelet packet transform, and extract the frequency domain features of each band respectively;

[0066] S50: Calculate the energy distribution of different frequencies in space using the frequency domain features of each channel, and extract the spatial features of each band.

[0067] In this embodiment, the S10 includes:

[0068] S11, using an optically pumped magnetometer to collect multi-channel magnetocardiographic signals, and filtering out invalid channels based on the correlation of the autocorrelation functions between the channels; when the average correlation coefficient between the autocorrelation functions of a certain channel and other channels is lower than a set threshold, the channel is considered to be an invalid channel and is removed;

[0069] S12, performing wavelet packet decomposition on the signal, setting the approximate coefficients representing baseline drift in the wavelet packet to zero; identifying the wavelet packet coefficients containing power frequency noise based on the kurtosis and kurtosis of the wavelet packet coefficients in the frequency domain, and setting the coefficients to zero; performing soft threshold denoising on all wavelet packet coefficients, and then reconstructing to obtain the denoised magnetocardiographic signal;

[0070] Wherein, the S11 includes:

[0071]

[0072] Among them, i represents the channel to be judged, j represents other channels except channel i, R i , R j represents the autocorrelation function between channel i and channel j, and L represents the total length of the acquired data;

[0073] The S12 includes:

[0074]

[0075] Among them, sig represents the denoised magnetic cardiogram signal, μ i represents the i-th wavelet packet coefficient after denoising.

[0076] In this embodiment, the S20 includes:

[0077] S21: Locating the R peak of the magnetocardiographic signal using a peak-finding algorithm;

[0078] S22: After locating the R peak, set the R wave to zero and perform peak location again. Find the P wave position at a sampling point before the Q wave, and find the T wave position at b sampling points after the Q wave, completing the location of the P, QRS, and T wave bands.

[0079] S23: Complete the P, QRS, and T band magnetic cardiogram signals u according to the set waveform width P，i (t),u R，i (t),u T，i (t) where i represents the i-th channel and t represents the t-th sampling point.

[0080] In this embodiment, the S30 includes:

[0081] Normalize the amplitude of each band to extract the parameters of each band. The waveform parameters include: Kurtosis, Skewness, Impulse Factor, Waveform Factor, Crest Factor and Margin Factor.

[0082] The feature parameters are combined into a time domain feature matrix ET of size 3*6 P , E.T. R , E.T. T , which consists of 6 waveform characteristic parameters of P, QRS, and T bands;

[0083] Among them, normalizing the amplitude of each band includes:

[0084]

[0085] Among them, u P ,u R ,u T represent the extracted P, QRS, and T wave bands, respectively;

[0086] Calculations of kurtosis, skewness, impulse factor, shape factor, crest factor, and margin factor include:

[0087]

[0088] in, represents the average value of the signal s(t), and L represents the total length of the acquired data.

[0089] In this embodiment, the S40 includes:

[0090] S41: Use wavelet packet transform to decompose the P, QRS, and T bands of each cardiac cycle of each channel respectively, and obtain the wavelet packet coefficient μ of each band P i，j , μ R i，j , μ T i，j , where i represents the i-th channel and j represents the j-th subband coefficient obtained by the 4th layer wavelet packet decomposition. P i，j , μ R i，j , μ T i，j The expressions are:

[0091]

[0092] Where h j (n) is the filter coefficient of the ith subband of the 4th layer, t is the index of the subband coefficient, 2 4 represents the sampling interval of the 4th layer subband, and n represents the index of the original signal.

[0093] S42: Based on the wavelet packet decomposition results, the energy weights of the wavelet packet coefficients in the P, QRS, and T bands are calculated respectively to obtain the energy distribution of each band in different frequency bands, including:

[0094]

[0095] Among them, EF P i,j , EF R i,j , EF Ti,j Respectively represent the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, μ Pi,j , μ Ri,j , μ Ti,j Represent the jth wavelet packet coefficient of the i-th channel of the P, QRS, and T bands respectively.

[0096] The obtained frequency domain parameters EFP i,j , EF R i,j , EF Ti,j Form three matrices EF of size N*M respectively P , EF R , EF T , respectively characterize the frequency domain characteristics of the P-band, QRS-band, and T-band of the single-channel magnetocardiographic signal. Where M represents the number of sensors, and N represents the number of wavelet packet coefficients obtained by decomposition.

[0097] In this embodiment, the S50 includes:

[0098] Using the multi-channel magnetic heart signal u(t), according to the arrangement of the sensor array, the energy distribution of the spatial magnetic field generated by the cardiac electrical activity is calculated in space, and the energy spatial distribution of different frequency bands in each band is obtained, that is, the spatial domain feature. The spatial domain feature consists of three N*M size feature matrices ES P , ES R , ES T Characterize, representing the spatial energy distribution of P wave, QRS wave and T wave respectively. Where M is the number of sensor channels and N is the number of wavelet packet coefficients. The specific calculation method is:

[0099]

[0100] ES P i,j , ES R i,j , ES T i,j Respectively represent the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel. P i,j , μ R i,j , μ T i,j represents the jth wavelet packet coefficient of the i-th channel of the P, QRS, and T bands, respectively, ||μ P i,j ||,||μ R i,j ||,||μ T i,j || represents the bi-norm of the j-th wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, respectively.

[0101] Example 2

[0102] This embodiment, based on the method provided in the first embodiment, proposes a multi-domain feature extraction method for multi-channel magnetic cardiogram signals of rats. The overall flow chart is as follows: Figure 1 As shown, specifically, the implementation process is as follows;

[0103] S10: A 9-channel rat cardiac magnetometry signal was acquired using a SERF magnetometer at a sampling rate of 1000 Hz. Invalid channels were screened based on inter-channel synchronization. The criterion for determining an invalid channel as the i-th channel was that the average correlation coefficient with the autocorrelation function of the other channels was lower than a set threshold of 0.2:

[0104]

[0105] Among them, i represents the channel to be judged, j represents other channels except channel i, R i , R j represents the autocorrelation function between channel i and channel j, and L represents the total length of the acquired data.

[0106] Taking a single channel as an example, to improve computational speed, the signal was first downsampled to a sampling rate of 500 Hz. Wavelet packet decomposition was then performed on the signal, using the Daubechies wavelet basis function and four decomposition layers to obtain 16 wavelet packet coefficients. The His transform was used to calculate the frequency domain envelope of each wavelet packet coefficient, and its kurtosis and kurtosis were calculated. When both kurtosis and kurtosis were greater than 0.8, the wavelet packet coefficient was considered to represent power frequency noise and was set to zero. Soft thresholding was then applied to all wavelet packet coefficients to denoise them, and the rat magnetocardiographic signal was reconstructed.

[0107]

[0108] Among them, sig represents the rat magnetic heart signal after noise reduction, μ i represents the i-th wavelet packet coefficient after denoising.

[0109] S20: Using a peak-finding algorithm, locate the R peak of the rat's magnetic cardiogram signal. After locating the R peak, reset the R wave to zero and perform peak-finding again. The P wave is located 100 sampling points before the R wave, and the T wave is located 150 sampling points after the R wave, completing the localization of the P, QRS, and T wave bands. The P wave width is set to 50 sampling points, the QRS wave width to 50 sampling points, and the T wave width to 60 sampling points to complete the division of each band.

[0110] S30: Normalize the amplitude of each band:

[0111]

[0112] Among them, u P ,u R ,u T Represent the extracted P, QRS, and T wave bands respectively.

[0113] Extract the parameters of each band, waveform parameters include: kurtosis, skewness, pulse factor, shape factor, peak factor and margin factor. The characteristic parameters are combined into a time domain characteristic matrix ET of size 3*6 P , E.T. R , E.T. T , which consists of 6 waveform characteristic parameters of P, QRS, and T bands;

[0114] S40: Use wavelet packet transform to decompose the P, QRS, and T bands of each cardiac cycle in each channel. Select the wavelet basis function as "Daubechies" and set the decomposition level to 4. Based on the wavelet packet decomposition results, calculate the energy weight of each wavelet packet coefficient in the P, QRS, and T bands to obtain the energy distribution of each band in different frequency bands. The energy weight calculation formula is:

[0115]

[0116] Among them, EF Pi,j ,EF Ri,j ,EF Ti,j Respectively represent the energy proportion of the i-th wavelet packet coefficient in the P band, QRS band, and T band in the i-th channel data, μ Pi,j , μ Ri,j , μ Ti,j represents the jth wavelet packet coefficient of the i-th channel of the P, QRS, and T bands, respectively, ||μ Pi,j ||,||μ Ri,j ||,||μ Ti,j || respectively represent the bi-norm of the j-th wavelet packet coefficient of the i-th channel of the P band.

[0117] For the multi-channel magnetic cardiogram signal of rats, the frequency domain parameter matrix of each channel is combined into three feature matrices EF of size 4*16 P , EF R , EF T , respectively characterizing the frequency domain characteristics of rat magnetic cardiac signals in each channel of the P band, QRS band, and T band.

[0118] S50: Utilizing the single-channel frequency domain characteristics and according to the arrangement of the rat sensor array, the energy distribution matrix of each frequency band at different bands is calculated to obtain the energy spatial distribution characteristics of each band of the rat.

[0119]

[0120] Among them, ES P i,j , ES R i,j , ES Ti,jRespectively represent the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel. Pi,j , μ Ri,j , μ Ti,j represents the jth wavelet packet coefficient of the i-th channel of the P, QRS, and T bands, respectively, ||μ Pi,j ||,||μ Ri,j ||,||μ Ti,j || represents the bi-norm of the j-th wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, respectively.

[0121] The spatial energy distribution matrix is 3*4*16 in size and consists of three 4*16 matrices, representing the spatial domain characteristics of the P wave, QRS wave, and T wave, respectively. 4 is the number of sensor channels, and 16 is the number of wavelet packet coefficients.

[0122] This example proposes a multi-domain feature extraction method for multi-channel cardiac magnetometry signals in rats. By combining waveform segmentation, the time domain features and frequency domain features of the P, QRS, and T waveforms of the rat cardiac magnetometry signals are extracted respectively. Based on the spatial magnetic field energy distribution, the unique spatial domain features of multi-channel cardiac magnetometry data are extracted. This method has important implications for rat model experiments based on cardiac magnetometry equipment and their data analysis.

[0123] Example 3

[0124] The present invention also provides a multi-domain feature extraction system for a multi-channel magnetocardiogram, the system being used to implement the aforementioned method, the system comprising: an acquisition module, a division module, an extraction module, a decomposition module, and a calculation module;

[0125] The acquisition module is used to collect multi-channel cardiac magnetometry signals through a cardiac magnetometry measurement device, check data quality, eliminate invalid channels, and remove power frequency interference and baseline drift in the cardiac magnetometry signals;

[0126] A division module is used to apply a waveform positioning algorithm to divide the magnetic cardiogram signal into the P wave band, the QRS wave band, and the T wave band;

[0127] An extraction module is used to extract the P-band, QRS-band, and T-band morphological parameters of the magnetic cardiogram signal as time-domain feature matrices;

[0128] The decomposition module is used to decompose the P-band, QRS-band, and T-band magnetic cardiogram signals using wavelet packet transform and extract the frequency domain features of each band respectively;

[0129] The calculation module is used to calculate the energy distribution of different frequencies in space by using the frequency domain characteristics of each channel and extract the spatial characteristics of each band.

[0130] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A multi-domain feature extraction method for multi-channel magnetocardiogram, characterized in that: The method comprises: S10: Collect multi-channel cardiac magnetometry signals through a cardiac magnetometry measurement device, check data quality, remove invalid channels, and remove power frequency interference and baseline drift in the cardiac magnetometry signals; S20: Apply waveform localization algorithm to divide the magnetic cardiogram signal into P wave, QRS wave and T wave; S30: extracting the P-band, QRS-band, and T-band morphological parameters of the magnetic cardiogram signal as time domain feature matrices; S40: Decompose the P-band, QRS-band, and T-band magnetocardiographic signals using wavelet packet transform, and extract the frequency domain features of each band respectively; S50: Calculate the energy distribution of different frequencies in space using the frequency domain features of each channel, and extract the spatial features of each band.

2. The method according to claim 1, characterized in that The S10 includes: S11, using an optically pumped magnetometer to collect multi-channel magnetocardiographic signals, and filtering out invalid channels based on the correlation of the autocorrelation functions between the channels; when the average correlation coefficient between the autocorrelation functions of a certain channel and other channels is lower than a set threshold, the channel is considered to be an invalid channel and is removed; S12, performing wavelet packet decomposition on the signal, setting the approximate coefficients representing baseline drift in the wavelet packet to zero; identifying the wavelet packet coefficients containing power frequency noise based on the kurtosis and kurtosis of the wavelet packet coefficients in the frequency domain, and setting the coefficients to zero; performing soft threshold denoising on all wavelet packet coefficients, and then reconstructing to obtain the denoised magnetocardiographic signal; Wherein, the S11 includes: Among them, i represents the channel to be judged, j represents other channels except channel i, R i , R j represents the autocorrelation function between channel i and channel j, and L represents the total length of the acquired data; The S12 includes: Among them, sig represents the denoised magnetic cardiogram signal, μ i represents the i-th wavelet packet coefficient after denoising.

3. The method according to claim 1, characterized in that The S20 includes: S21: Locating the R peak of the magnetocardiographic signal using a peak-finding algorithm; S22: After locating the R peak, set the R wave to zero and perform peak location again. Find the P wave position at a sampling point before the Q wave, and find the T wave position at b sampling points after the Q wave, completing the location of the P, QRS, and T wave bands. S23: Complete the P, QRS, and T band magnetic cardiogram signals u according to the set waveform width P,i (t),u R，i (t),u P,i (t), where i represents the i-th channel and t represents the t-th sampling point.

4. The method according to claim 1, wherein The S30 includes: Normalize the amplitude of each band to extract the parameters of each band, including kurtosis, skewness, pulse factor, shape factor, peak factor and margin factor; The feature parameters are combined into a time domain feature matrix ET of size 3*6 P ,ET R ,ET T , which consists of 6 waveform characteristic parameters of P, QRS, and T bands; Among them, normalizing the amplitude of each band includes: Among them, u P ,u R ,u T represent the extracted P, QRS, and T wave bands, respectively; Calculations of kurtosis, skewness, impulse factor, shape factor, crest factor, and margin factor include: in, represents the average value of the signal s(t), and L represents the total length of the acquired data.

5. The method according to claim 1, wherein The S40 includes: S41: using wavelet packet transform to decompose the P, QRS, and T bands of each cardiac cycle of each channel respectively, and obtain the wavelet packet coefficients of each band; S42: Based on the wavelet packet decomposition results, the energy weights of the wavelet packet coefficients in the P, QRS, and T bands are calculated respectively to obtain the energy distribution of each band in different frequency bands; the obtained frequency domain parameters are combined into a matrix to characterize the frequency domain characteristics of the single-channel magnetocardiographic signal; Among them, according to the wavelet packet decomposition results, the energy weights of each wavelet packet coefficient in the P, QRS, and T bands are calculated respectively, and the energy distribution of each band in different frequency bands is obtained, including: Among them, EF Pi,j , EF Ri,j , EF Ti,j Respectively represent the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, μ Pi,j , μ Ri,j , μ Ti,j represents the jth wavelet packet coefficient of the i-th channel of the P, QRS, and T bands respectively; The obtained frequency domain parameters are combined into a matrix to characterize the frequency domain characteristics of the single-channel magnetic cardiogram signal, including: The obtained frequency domain parameters EF Pi,j , EF Ri,j , EF Ti,j Form three matrices ET of size N*M respectively P ,ET R ,ET T , respectively characterizing the frequency domain characteristics of the P band, QRS band and T band of the single-channel magnetic cardiogram signal, where M represents the number of sensors and N represents the number of wavelet packet coefficients obtained by decomposition.

6. The method according to claim 1, characterized in that The S50 includes: Using multi-channel magnetocardiographic signals and the sensor array arrangement, the energy distribution matrix of each frequency band at different bands is calculated to obtain the energy spatial distribution of each band, that is, the spatial domain characteristics; The spatial domain features are composed of three N*M matrices, representing the spatial domain features of P wave, QRS wave, and T wave respectively, where N is the number of sensor channels and M is the number of wavelet packet coefficients; Among them, the energy distribution matrix of each frequency band at different bands is calculated to obtain the energy spatial distribution of each band, that is, the spatial domain characteristics, including: Among them, ES P i,j , ES R i,j , ES Ti,j represents the spatial energy proportion of the jth wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, respectively, ||μ P i,j ||,||μ R i,j ||,||μ T i,j || represents the bi-norm of the j-th wavelet packet coefficient in the P, QRS, and T bands in the i-th channel, respectively.

7. A multi-domain feature extraction system for a multi-channel magnetocardiogram, the system being used to implement the method according to any one of claims 1 to 6, characterized in that: The system includes: an acquisition module, a division module, an extraction module, a decomposition module and a calculation module; The acquisition module is used to collect multi-channel cardiac magnetometry signals through a cardiac magnetometry measurement device, check data quality and eliminate invalid channels, and remove power frequency interference and baseline drift in the cardiac magnetometry signals; The division module is used to apply a waveform positioning algorithm to divide the magnetic cardiogram signal into a P wave band, a QRS wave band, and a T wave band; The extraction module is used to extract the P-band, QRS-band, and T-band morphological parameters of the magnetocardiographic signal as time domain feature matrices; The decomposition module is used to decompose the P-band, QRS-band, and T-band magnetocardiographic signals using wavelet packet transform to extract frequency domain features of each band; The calculation module is used to calculate the energy distribution of different frequencies in space by using the frequency domain characteristics of each channel and extract the spatial characteristics of each band.

Images