Multi-channel joint time-space domain filtering method for magnetocardiogram signals
Through the multi-channel combined space-time filtering method, Hermite fitting and clustering technology is used to solve the problems of poor cardiac image quality and poor noise suppression effect under single-channel processing, and efficient noise reduction and signal quality of cardiac signal are achieved promote.
Patent Information
- Application Number
- CN202410996745.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-24
AI Technical Summary
The existing core magnetic filtering methods are mainly single-channel processing, and cannot effectively utilize the correlation between multiple channels, resulting in poor quality of core magnetic image and poor non-Gaussian noise suppression effect on spectrum overlap.
Using multi-channel combined space-time filtering method, the core magnetic signal is processed through Hermite fitting and clustering technology. Using the high synchronization and repetition characteristics of multi-channel signals, noise reduction is carried out in the time and space domains to build different A set of characteristic waveform cardiac image blocks and a weighted average of the signals.
The suppression effect of spectral overlapping non-Gaussian noise is improved, the signal-to-noise ratio of the magnetic core signal is improved, the spatial characteristics of the magnetic core image are retained, and the quality of the magnetic core image is significantly improved.
Smart Images

Figure CN118941461A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of magnetocardiographic signal processing, and in particular relates to a multi-channel joint spatiotemporal filtering method for magnetocardiographic signals. Background Art
[0002] Cardiovascular disease poses a persistent threat to human life, health and safety. More than four-fifths of the risk of premature heart disease and stroke can be avoided through reasonable means. Therefore, it is of great significance to improve the convenience and accuracy of cardiovascular disease monitoring and diagnosis. Magnetocardiography, as a non-invasive measurement of cardiac electrophysiological activity, can study the state of the heart. Compared with traditional electrocardiogram measurement devices, extremely weak magnetocardiography devices have the advantages of non-contact measurement, high sensitivity and strong specificity. Preoperative magnetocardiography can provide valuable information for medical treatment.
[0003] The magnetocardiogram signal is extremely weak. In actual detection, it is easily affected by environmental noise, instrument noise, human eye movement and breathing noise, especially non-Gaussian noise overlapping with the signal spectrum, which seriously affects the signal-to-noise ratio. As a spatial signal, the magnetocardiogram signal itself contains rich spatial information. The current magnetocardiogram filtering method mainly uses a single-channel processing method, which ignores the correlation between each channel, cannot effectively guarantee the quality of the magnetocardiogram, and has poor suppression effect on non-Gaussian noise with overlapping spectrum. Therefore, how to use the temporal and spatial information of the magnetocardiogram to improve the quality of the magnetocardiogram is of great significance in this field. Summary of the invention
[0004] In order to solve the technical problems of poor quality of magnetocardiographic images and poor suppression effect of non-Gaussian noise with spectral overlap in existing magnetocardiographic filtering methods mainly based on single-channel processing, the present invention provides a multi-channel joint spatiotemporal filtering method for magnetocardiographic signals.
[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:
[0006] A multi-channel joint spatiotemporal filtering method for magnetocardiographic signals of the present invention comprises the following steps:
[0007] Step 1: Collect multi-channel magnetic cardiograph signals through a magnetocardiograph;
[0008] Step 2: Based on the multi-channel magnetic cardiology signal, use the Hermite fitting method to fit each channel to obtain the Hermite fitting parameters of each channel at each moment;
[0009] Step 3: Divide the magnetocardiographic image into several blocks of size n*n in space, and use the Hermite fitting parameters of the corresponding channel to represent the magnetocardiographic image of each block at each moment;
[0010] Step 4: Based on the Hermite fitting parameters, clustering method is used to cluster all the magnetocardiographic image blocks at all times to construct a set of magnetocardiographic image blocks with different characteristic waveforms;
[0011] Step 5: within the range of the magnetocardiographic image block set with different characteristic waveforms, the weight of the magnetocardiographic signal of each magnetocardiographic image block is calculated according to the Hermite fitting parameters, and the magnetocardiographic signal of each magnetocardiographic image block is weighted averaged according to the calculated weight information to obtain the denoised magnetocardiographic signal.
[0012] Furthermore, each sampling point is represented by Hermite fitting parameters that characterize the waveform characteristics.
[0013] Furthermore, the Hermite fitting parameters are fitted using the least squares method, and the specific fitting formula is as follows:
[0014]
[0015] Where σ represents the time scale factor of the Hermite fitting, c n represents the nth order Hermite coefficient obtained by Hermite fitting of the data segment, x(t) represents the data segment to be fitted, t∈T represents the tth sampling point of the data segment, Φ n (t,σ) represents the n-order Hermite basis function, H n (t / σ) represents the nth-order Hermite polynomial with t / σ as the independent variable. The Hermite polynomial is obtained by a recursive formula. In the recursive formula, H1(x) represents the first-order Hermite polynomial, H n (x) represents the Nth order Hermite polynomial, H n-1 (x) represents the n-1th order Hermite polynomial, H n-2 (x) represents the n-2th order Hermite polynomial.
[0016] Furthermore, in step 3, the number of magnetocardiographic image blocks obtained is Ch / n*n*T, where Ch represents the number of channels and T represents the number of magnetocardiographic signal samples.
[0017] Furthermore, in step 4, a clustering method is used to cluster all magnetocardiographic image blocks at all times according to the low-order Hermite coefficient of the n*n block at each moment, and K cluster centers are found through iteration, and all magnetocardiographic image block data are divided into K sets.
[0018] Furthermore, in step 4, the clustering method adopts K-means algorithm.
[0019] Furthermore, in step 5, the weight of the magnetocardiographic signal of each magnetocardiographic image block is derived from the similarity coefficient between each magnetocardiographic image block calculated by Hermite fitting.
[0020] Furthermore, in step 5, the specific calculation formula of the weight of the magnetocardiographic signal of each magnetocardiographic image block is as follows:
[0021]
[0022] Where w(i, j) represents the weight of the jth sampling point to the ith sampling point, i.e., the similarity coefficient, and Z represents the normalization coefficient, ensuring c i,t represents the t-order Hermite coefficient of the i-th sampling point, c j,t represents the t-order Hermite coefficient of the j-th sampling point, h k The sum of the variances of the Hermite coefficients of all sampling points in the kth set obtained by clustering is the filter coefficient, M represents the total Hermite order, and n represents the side length of the magnetocardiographic image block.
[0023] Furthermore, the filter coefficient h k The specific calculation formula is as follows:
[0024]
[0025] Where N is the number of elements in the set, M is the total Hermite order, n is the side length of the magnetocardiographic image block, and c is n,m represents the m-th order Hermite coefficient of the n-th sampling point, Represents the mean of the m-th order Hermite coefficients of all N elements in a cluster set.
[0026] The beneficial effects of the present invention are:
[0027] The invention discloses a multi-channel joint spatiotemporal filtering method for a magnetocardiogram signal. First, a magnetocardiogram instrument is used to collect multi-channel magnetocardiogram signals, and all channels are fitted by the Hermite fitting method to obtain the Hermite fitting parameters of each channel at each moment. Then, the magnetocardiogram image is spatially divided into a plurality of blocks, and the Hermite fitting parameters of the corresponding channels are used to represent the magnetocardiogram image at each moment. Then, all blocks at all moments are clustered to construct a magnetocardiogram block set with different characteristic waveforms. Finally, a weighted average of the magnetocardiogram signal is performed within the range of the magnetocardiogram block set with different characteristic waveforms to obtain a denoised magnetocardiogram signal. The weight is derived from the similarity between the magnetocardiogram blocks calculated by the Hermite fitting.
[0028] The multi-channel joint spatiotemporal filtering method of the magnetocardiogram signals of the present invention utilizes the high synchronization and high repeatability of the multi-channel magnetocardiogram signals to perform noise reduction processing on the magnetocardiogram images in the time domain and the space domain, and can adaptively process a variety of non-Gaussian noises, thereby improving the suppression effect of non-Gaussian noise with spectral overlap. The combination of time denoising and spatial denoising is helpful to improve the quality of two-dimensional magnetocardiogram data, while retaining the time domain characteristics of the magnetocardiogram signals, improving the signal-to-noise ratio of the magnetocardiogram signals, and protecting the spatial characteristics of the magnetocardiogram images while improving the signal-to-noise ratio of the single-channel signal. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 A schematic flow chart of a multi-channel joint spatiotemporal filtering method for magnetocardiographic signals provided by the present invention. DETAILED DESCRIPTION
[0030] The present invention provides a multi-channel joint time-space domain filtering method for magnetocardiogram, which reduces noise of magnetocardiogram signals from the time domain and space domain of magnetocardiogram, and improves the quality of magnetocardiogram by using the time and space information of magnetocardiogram.
[0031] To achieve the above object, the present invention provides a multi-channel joint spatiotemporal filtering method for magnetocardiographic signals. Figure 1 The present invention specifically comprises the following steps:
[0032] Step 1: Collect the multi-channel magnetocardiogram signals of the tested person through a magnetocardiograph, check the data quality, remove invalid or abnormal channels, record the number of channels as Ch, and the number of magnetocardiogram signal samples as T, and finally obtain a noisy multi-channel magnetocardiogram signal.
[0033] The number of channels Ch is preferably 64, the sampling rate is preferably 1000 Hz, and the acquisition time is preferably 60 seconds. During the acquisition process, the person being tested is required not to carry any metal objects to avoid interference with the magnetic cardiology equipment.
[0034] Step 2: Based on the multi-channel magnetocardiographic signal, use the Hermite fitting method to fit each channel to obtain the Hermite fitting parameters of each channel at each moment, and use the Hermite fitting parameters that characterize the waveform characteristics to represent each sampling point.
[0035] The Hermite fitting parameters of each channel at each moment can be obtained by fitting all channels using the Hermite fitting method, wherein the length of the data to be fitted is preferably a data segment of 100ms in length, and the Hermite polynomial order is preferably 6. The time domain signal of each channel of the noisy magnetic cardiogram signal at each moment is represented by a low-order Hermite coefficient obtained based on the Hermite fitting, that is, each data segment of 100ms in length is compressed into a 6th-order Hermite coefficient.
[0036] Among them, the Hermite fitting parameters can be fitted using the least squares method, and the specific solution steps are as follows:
[0037] Extract the data segment of 100ms length near the sampling point s, fit the M-order Hermite coefficient in the range near the single-channel time domain signal through the following formula, and compress the sampling points within the length of 100ms into a vector of length M. The specific fitting formula is as follows:
[0038]
[0039] Where σ represents the time scale factor of the Hermite fitting, c n represents the nth order Hermite coefficient obtained by Hermite fitting of the data segment, x(t) represents the data segment to be fitted, t∈T represents the tth sampling point of the data segment, Φ n (t,σ) represents the n-order Hermite basis function, H n (t / σ) represents the nth-order Hermite polynomial with t / σ as the independent variable. The Hermite polynomial can be obtained by the recursive formula. In the recursive formula, H1(x) represents the first-order Hermite polynomial, H n (x) represents the Nth order Hermite polynomial, H n-1 (x) represents the n-1th order Hermite polynomial, H n-2 (x) represents the n-2th order Hermite polynomial;
[0040] Step 3: Divide the magnetocardiographic image into several blocks of size n*n in space, and use the Hermite fitting parameters of the corresponding channels to represent the magnetocardiographic image of each block at each moment. The number of magnetocardiographic image blocks is Ch / n*n*T, where Ch represents the number of channels and T represents the number of magnetocardiographic signal samples.
[0041] Step 4: Based on the Hermite fitting parameters, a clustering method is used to cluster all the magnetocardiographic image blocks at all times to construct a set of magnetocardiographic image blocks with different characteristic waveforms.
[0042] Among them, a clustering method is used to cluster all magnetocardiographic image blocks at all times according to the low-order Hermite coefficient of the n*n block at each moment, and K cluster centers are found through iteration, and all magnetocardiographic image block data are divided into K sets.
[0043] Specifically, n=2, K=50, that is, each frame of the magnetocardiographic image is spatially divided into several 2*2 blocks, and the four 6th-order Hermite coefficients are merged to represent the 2*2 blocks of each frame as a 4*6 Hermite coefficient matrix. According to the low-order Hermite coefficients of the 2*2 blocks at each moment, the K-means algorithm is used to cluster all magnetocardiographic image blocks at all moments, and 50 cluster centers are found through iteration, and all magnetocardiographic image block data are divided into 50 sets.
[0044] It should be noted that the K-means algorithm is an existing technology well known to professionals in the field, so it will not be described in detail; for clustering of magnetocardiographic image blocks, in addition to using the maximum likelihood estimation method, any existing method can also be used, and the present invention does not specifically limit the estimation method.
[0045] Step 5: Within the range of the magnetocardiographic image block set with different characteristic waveforms, the weight of the magnetocardiographic signal of each magnetocardiographic image block is calculated according to the Hermite fitting parameters, and the weighted average of the magnetocardiographic signal of each magnetocardiographic image block is performed according to the calculated weight information to obtain the denoised magnetocardiographic signal. The weight is derived from the similarity coefficient between each magnetocardiographic image block calculated by the Hermite fitting.
[0046] The specific calculation formula of the similarity coefficient between each magnetocardiographic image block is as follows:
[0047]
[0048] Among them, ω(i,j) represents the weight of the jth sampling point to the ith sampling point, that is, the similarity coefficient, and Z represents the normalization coefficient to ensure c i,t represents the t-order Hermite coefficient of the i-th sampling point, c j,t represents the t-order Hermite coefficient of the j-th sampling point, h k The sum of the variances of the Hermite coefficients of all sampling points in the kth set obtained by clustering is the filter coefficient, M represents the total Hermite order, and n represents the side length of the magnetocardiographic image block.
[0049] Among them, the following formula can be used as the filter coefficient h k Global adaptation of:
[0050]
[0051] Where N is the number of elements in the set, M is the total Hermite order, n is the side length of the magnetocardiographic image block, and c is n,m represents the m-th order Hermite coefficient of the n-th sampling point, Represents the mean of the m-th order Hermite coefficients of all N elements in a cluster set.
[0052] In summary, the present invention provides a multi-channel joint spatiotemporal filtering method for a magnetocardiogram signal, which collects multi-channel magnetocardiogram signals through a magnetocardiograph, fits all channels using the Hermite fitting method, obtains the Hermite fitting parameters of each channel at each moment, divides the magnetocardiogram image into several blocks in space, and uses the Hermite fitting parameters of the corresponding channels to represent the magnetocardiogram image at each moment, clusters all blocks at all moments, constructs a set of magnetocardiogram image blocks with different characteristic waveforms, and weights the magnetocardiogram signal within the range of the magnetocardiogram image block set with different characteristic waveforms, and the weights are derived from the similarities between the magnetocardiogram image blocks calculated by the Hermite fitting. The present invention utilizes the high synchronization and high repeatability of multi-channel magnetocardiogram signals, reduces the noise of the magnetocardiogram image in the time domain and the space domain, and helps to protect the spatial characteristics of the magnetocardiogram image while improving the signal-to-noise ratio of the single-channel signal.
[0053] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed in the present invention shall still be covered by the protection scope of the present invention.
Claims
1. A multi-channel joint spatiotemporal filtering method for magnetocardiographic signals, characterized in that: The following steps are involved: Step 1: Collect multi-channel magnetic cardiograph signals through a magnetocardiograph; Step 2: Based on the multi-channel magnetic cardiology signal, use the Hermite fitting method to fit each channel to obtain the Hermite fitting parameters of each channel at each moment; Step 3: Divide the magnetocardiographic image into several blocks of size n*n in space, and use the Hermite fitting parameters of the corresponding channel to represent the magnetocardiographic image of each block at each moment; Step 4: Based on the Hermite fitting parameters, clustering method is used to cluster all the magnetocardiographic image blocks at all times to construct a set of magnetocardiographic image blocks with different characteristic waveforms; Step 5: within the range of the magnetocardiographic image block set with different characteristic waveforms, the weight of the magnetocardiographic signal of each magnetocardiographic image block is calculated according to the Hermite fitting parameters, and the magnetocardiographic signal of each magnetocardiographic image block is weighted averaged according to the calculated weight information to obtain the denoised magnetocardiographic signal.
2. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: Each sampling point is represented by Hermite fitting parameters that characterize the waveform.
3. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: The Hermite fitting parameters are fitted using the least squares method, and the specific fitting formula is as follows: c n (s)-S t x(t)Φ n (t,σ); Where σ represents the time scale factor of the Hermite fitting, c n represents the nth order Hermite coefficient obtained by Hermite fitting of the data segment, x(t) represents the data segment to be fitted, t∈T represents the tth sampling point of the data segment, Φ n (t,σ) represents the n-order Hermite basis function, H n (t / σ) represents the nth-order Hermite polynomial with t / σ as the independent variable. The Hermite polynomial is obtained by a recursive formula. In the recursive formula, H1(x) represents the first-order Hermite polynomial, H n (x) represents the Nth order Hermite polynomial, H n-1 (x) represents the n-1th order Hermite polynomial, H n-2 (x) represents the n-2th order Hermite polynomial.
4. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: In step 3, the number of magnetocardiographic image blocks obtained is Ch / n*n*T, where Ch represents the number of channels and T represents the number of magnetocardiographic signal samples.
5. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: In step 4, a clustering method is used to cluster all magnetocardiographic image blocks at all times according to the low-order Hermite coefficient of the n*n block at each moment, and K cluster centers are found through iteration, and all magnetocardiographic image block data are divided into K sets.
6. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: In step 4, the clustering method adopts K-means algorithm.
7. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: In step 5, the weight of the magnetocardiographic signal of each magnetocardiographic image block is derived from the similarity coefficient between each magnetocardiographic image block calculated by Hermite fitting.
8. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 1, characterized in that: In step 5, the specific calculation formula of the weight of the magnetocardiographic signal of each magnetocardiographic image block is as follows: Where w(i, j) represents the weight of the jth sampling point to the ith sampling point, i.e., the similarity coefficient, and Z represents the normalization coefficient, ensuring c i,t represents the t-order Hermite coefficient of the i-th sampling point, c j,t represents the t-order Hermite coefficient of the j-th sampling point, h k The sum of the variances of the Hermite coefficients of all sampling points in the kth set obtained by clustering is the filter coefficient, M represents the total Hermite order, and n represents the side length of the magnetocardiographic image block.
9. The multi-channel joint spatiotemporal filtering method of magnetocardiographic signals according to claim 8, characterized in that: The filter coefficient h k The specific calculation formula is as follows: Where N is the number of elements in the set, M is the total Hermite order, n is the side length of the magnetocardiographic image block, and c is n,m represents the m-th order Hermite coefficient of the n-th sampling point, Represents the mean of the m-th order Hermite coefficients of all N elements in a cluster set.
Citation Information
Patent Citations
Efficient and accurate heart magnetic field measurement and analysis method
CN118000740A
Filtering device, method, program, and recording medium
WO2021181969A1
