Muscle magnetic signal denoising method based on improved variational mode decomposition and soft threshold processing
By improving VMD and soft thresholding methods, the problem of poor noise removal in MRI signals was solved, achieving efficient denoising of MRI signals, adapting to their non-stationary characteristics and improving signal quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-28
- Publication Date
- 2026-04-10
AI Technical Summary
Existing technologies are not effective in removing noise from myomagnetic signals, especially when faced with complex noise. Traditional methods such as IIR filters and EMD have limitations and cannot effectively remove various noise interferences.
An improved variational mode decomposition (VMD) combined with a soft thresholding method is adopted. The myomagnetic signal is decomposed into multiple variational mode functions (VMF) by VMD, and noise is identified and removed by frequency analysis. The component types are determined by fine composite multiscale scatter entropy (RCMDE) decomposition, and wavelet soft thresholding is used for noise reduction. Finally, the signal is reconstructed by linear superposition.
It effectively adapts to the non-stationary characteristics of myomagnetic signals, improves noise removal efficiency, accurately assesses signal complexity, and ensures signal quality.
Smart Images
Figure CN121834124A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of biomedical signal processing technology, and in particular to a method for denoising myoma signals based on improved variational mode decomposition and soft thresholding. Background Technology
[0002] Magnetomyography (MMG) is a non-invasive technique for measuring the magnetic signals of muscle activity. Compared to traditional electromyography (EMG), it offers advantages such as high spatial resolution and independence from tissue conductivity. However, MMG signals are extremely weak, typically on the order of picoteslas (pT), and exhibit wide bandwidth (10-200 Hz), nonlinearity, and non-stationarity. During acquisition, they are susceptible to various noise contaminations, including electric field interference, Gaussian white noise, and baseline drift, which severely affect subsequent analysis and application results.
[0003] Traditional electromyography (EMG) signal preprocessing methods, such as Infinite Impulse Response (IIR) filters and Empirical Mode Decomposition (EMD), have the following limitations in removing MMG noise: IIR struggles to remove noise overlapping with the signal spectrum; EMD is prone to mode aliasing, and several improved EMD methods suffer from excessive computational cost. Variational Mode Decomposition (VMD), as an adaptive non-recursive signal decomposition method, can decompose the input signal into multiple mode functions with finite bandwidth, overcoming mode aliasing and endpoint effects, and is more robust than traditional methods in feature extraction and noise reduction.
[0004] However, VMD alone cannot achieve complete denoising; modality selection and subsequent thresholding are significant challenges in VMD denoising. Furthermore, related techniques rely on the quality of the original signal for denoising and are insufficient when dealing with complex noise. Summary of the Invention
[0005] In view of this, the present application provides a method for denoising myoma signals based on improved variational mode decomposition and soft thresholding, in order to solve the problem of poor noise suppression effect of weak non-stationary myoma signals in the prior art.
[0006] A first aspect of this application provides a method for denoising myoma signals based on improved variational mode decomposition and soft thresholding, comprising:
[0007] Obtain the raw myoma magnetic resonance (MMG) signal; the raw MMG signal is a noisy signal.
[0008] Variational mode decomposition (VMD) is performed on the original MMG signal to obtain K variational mode functions (VMFs); K is a positive integer.
[0009] Frequency analysis is performed on K VMFs to identify the target VMF corresponding to the first type of noise. Noise removal processing is then performed based on the target VMF to obtain the VMF after the first denoising process.
[0010] The VMF after the first denoising process is decomposed using Refined Composite Multiscale Dispersion Entropy (RCMDE) to determine the component types of each VMF; the component types include noise component, noise-dominant component, signal component, and signal-dominant component.
[0011] The signal component VMF is preserved while the noise component VMF is removed. After local mean decomposition, wavelet soft thresholding is used to denoise both the signal-dominant VMF and the noise-dominant VMF.
[0012] The VMFs after the second denoising process are linearly superimposed to reconstruct the denoised MMG signal.
[0013] A second aspect of this application provides a myoma signal denoising device based on improved variational mode decomposition and soft thresholding, comprising:
[0014] The acquisition module is configured to acquire the raw myoma magnetic resonance (MMG) signal; the raw MMG signal is a noisy signal.
[0015] The first decomposition module is configured to perform variational mode decomposition (VMD) on the original MMG signal to obtain K variational mode functions (VMF); K is a positive integer.
[0016] The first denoising module is configured to perform frequency analysis on K VMFs, identify the target VMF corresponding to the first type of noise, and perform noise removal processing based on the target VMF to obtain the VMF after the first denoising process.
[0017] The second decomposition module is configured to decompose the VMF after the first denoising process using the Refined Composite Multiscale Spread Entropy (RCMDE) to determine the component types of each VMF; wherein, the component types include noise components, noise-dominant components, signal components, and signal-dominant components.
[0018] The second denoising module is configured to retain the signal component VMF and remove the noise component VMF, and to perform denoising processing on the signal dominant component VMF and the noise dominant component VMF after local mean decomposition using wavelet soft thresholding.
[0019] The reconstruction module is configured to linearly superimpose the VMFs after the second denoising process to reconstruct the denoised MMG signal.
[0020] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.
[0021] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.
[0022] The beneficial effects of the embodiments in this application compared with the prior art are:
[0023] This application embodiment obtains multiple VMFs by performing VMD processing on the original MMG signal. Frequency analysis is then performed on these VMFs to identify the target VMF corresponding to the first type of noise. The target VMF is then denoised to obtain the VMF after the first denoising process. RCMDE is then used to decompose the VMF after the first denoising process to determine the component type of each VMF. The signal component VMF is retained while the noise component VMF is removed. Simultaneously, the signal-dominant component VMF and the noise-dominant component VMF are denoised using wavelet soft thresholding after local mean decomposition. Finally, the VMFs after the second denoising process are linearly superimposed to reconstruct the denoised MMG signal. This method can effectively adapt to the non-stationary characteristics of MMG signals and accurately assess signal complexity, thereby improving the noise removal efficiency of MMG signals. Attached Figure Description
[0024] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0025] Figure 1 This is a flowchart illustrating a method for denoising myoma signals based on improved variational mode decomposition and soft thresholding, provided in an embodiment of this application.
[0026] Figure 2 This is a flowchart illustrating the method for decomposing the target VMF after the first denoising process using RCMDE, as provided in an embodiment of this application.
[0027] Figure 3This is a flowchart illustrating another method for denoising myoma signals based on improved variational mode decomposition and soft thresholding provided in this application embodiment.
[0028] Figure 4 This is a schematic diagram of a myoma signal denoising device based on improved variational mode decomposition and soft thresholding provided in an embodiment of this application.
[0029] Figure 5 This is a schematic diagram of the electronic device provided in the embodiments of this application. Detailed Implementation
[0030] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0031] The following will describe in detail, with reference to the accompanying drawings, a method and apparatus for denoising myoma signals based on improved variational mode decomposition and soft thresholding according to embodiments of this application.
[0032] As mentioned above, VMD alone cannot achieve complete denoising; modality selection and subsequent thresholding are important issues in VMD denoising. Furthermore, related technologies rely on the quality of the original signal during denoising and are insufficient when dealing with complex noise.
[0033] In view of this, this application provides a method for denoising MMG signals based on improved variational mode decomposition and soft thresholding. Addressing the lack of an effective mode classification and screening mechanism after traditional VMD decomposition, a mode screening framework is designed. By introducing RCMDE mode screening, component type identification of the VMF is achieved, effectively separating noise from the signal. Simultaneously, different processing strategies are adopted for modes containing different noise types. For baseline drift noise and power line interference noise, direct filtering or abandonment of the mode is used. For Gaussian white noise, a combination of LMD decomposition and wavelet soft thresholding denoising is used. This effectively removes various noises from the MMG signal, improves denoising efficiency, and avoids noise residue.
[0034] Figure 1 This is a schematic flowchart illustrating a method for denoising magnetic resonance imaging (MRI) signals based on improved variational mode decomposition and soft thresholding, as provided in an embodiment of this application. Figure 1 As shown, the method includes the following steps:
[0035] In step S101, the raw myoma magnetic signal is acquired.
[0036] Among them, the original myoma magnetic signal is a noisy signal.
[0037] In step S102, variational mode decomposition is performed on the original myoma magnetic signal to obtain K variational mode functions.
[0038] Where K is a positive integer.
[0039] In step S103, frequency analysis is performed on the K variational mode functions to identify the target variational mode function corresponding to the first type of noise. Noise removal processing is performed based on the target variational mode function to obtain the variational mode function after the first denoising process.
[0040] In step S104, the variational mode functions after the first denoising process are decomposed using fine composite multiscale scattering entropy to determine the component types of each variational mode function.
[0041] The component types include noise components, noise-dominant components, signal components, and signal-dominant components.
[0042] In step S105, the variational mode functions of the signal components are retained while the variational mode functions of the noise components are removed. After local mean decomposition, wavelet soft thresholding is used to denoise the variational mode functions of the dominant signal components and the variational mode functions of the dominant noise components.
[0043] In step S106, the variational mode functions after the second denoising process are linearly superimposed to reconstruct the denoised myoma signal.
[0044] In some embodiments of this application, the method may be executed by a server or by a terminal device with certain processing capabilities.
[0045] In some embodiments of this application, the original MMG signal can be obtained, which is a noisy signal.
[0046] The original MMG signal can then be processed by VMD to obtain K VMFs. VMD achieves signal decomposition by solving a constrained variational problem, which can be expressed as:
[0047] ;
[0048] in, It is a minimum value function. The L2 norm symbol, This represents the k-th VMF. The center frequency corresponding to the k-th VMF, This represents the partial derivative with respect to time t. It is the Dirac function, which represents the impulse response in signal processing. This is the original single-channel MMG signal. The imaginary unit satisfies , This indicates a constraint.
[0049] Furthermore, the mathematical model for iteratively solving variational problems can be: ; ;as well as ;in, and The first The modal function at the th ... Second and third Fourier transform at the next iteration The Fourier transform of the original single-channel MMG signal. For the first The modal function at the th ... Fourier transform at the next iteration For the Lagrange multipliers in the 1st Fourier transform at the next iteration This is a penalty factor used to ensure convergence. and The first The modal function at the th ... Second and third The center frequency at the next iteration and The Lagrange multipliers are respectively in the th Second and third Fourier transform at the next iteration This is the step size parameter, used to control the magnitude of the multiplier update.
[0050] The number of VMFs, K, is the number of VMD modes, which can be set according to actual needs. In one example, K=12 can be set.
[0051] In some embodiments of this application, frequency analysis can be performed on the K VMFs obtained by decomposition to identify the target VMF corresponding to the first type of noise, and noise removal processing can be performed based on the target VMF to obtain the VMF after the first denoising process.
[0052] In some embodiments of this application, RCMDE can also be used to decompose the VMF after the first denoising process to determine the component type of each VMF. The component type can include noise components, noise-dominant components, signal components, and signal-dominant components.
[0053] In some embodiments of this application, the signal component VMF can be retained while the noise component VMF is removed. After local mean decomposition, wavelet soft thresholding is used to denoise both the signal-dominant VMF and the noise-dominant VMF. Finally, the VMFs after the second denoising process are linearly superimposed to reconstruct the denoised MMG signal.
[0054] According to the technical solution provided in the embodiments of this application, multiple VMFs are obtained by performing VMD processing on the original MMG signal. Frequency analysis is then performed on these multiple VMFs to identify the target VMF corresponding to the first type of noise. The target VMF is then denoised to obtain the VMF after the first denoising process. Then, RCMDE is used to decompose the VMF after the first denoising process to determine the component type of each VMF. The signal component VMF is retained while the noise component VMF is removed. At the same time, the signal dominant component VMF and the noise dominant component VMF are denoised using wavelet soft thresholding after local mean decomposition. Finally, the VMFs after the second denoising process are linearly superimposed to reconstruct the denoised MMG signal. This method can effectively adapt to the non-stationary characteristics of MMG signals and accurately assess signal complexity, thereby improving the noise removal efficiency of MMG signals.
[0055] In some embodiments of this application, the first type of noise includes at least baseline drift noise. In this case, frequency analysis is performed on the K VMFs to identify the target VMF corresponding to the first type of noise. Noise removal based on the target VMF can be performed by first locating the baseline drift noise based on the characteristic that the center frequencies of the K VMFs are sorted from low to high, where the VMF containing the baseline drift noise is the first VMF; then, the first VMF is discarded, thus removing the baseline drift noise.
[0056] In other words, since baseline drift is a superposition of DC component and low-frequency signal, motion artifacts outside the experimental paradigm often cause baseline drift in MMG acquisition. As narrowband noise, its frequency is usually below 10 Hz. Therefore, based on the inherent characteristics of VMD, the K VMFs obtained after VMD decomposition are sorted by center frequency from low to high. Among them, the first VMF (i.e., the VMF with the lowest center frequency) usually contains baseline drift noise. Therefore, the first VMF can be directly discarded to remove baseline drift noise.
[0057] After the initial removal of baseline drift noise, the remaining VMFs can be low-pass filtered for subsequent processing. The low-pass filtering operation can be expressed by the following formula:
[0058] ;
[0059] in, The first one after low-pass filtering VMF, This is a low-pass filter function. For the first The frequency corresponding to the maximum amplitude of each VMF in the frequency domain. As the upper limit of the expected baseline drift, one example can be set to... .
[0060] In some embodiments of this application, the first type of noise may further include power line interference noise. In this case, frequency analysis is performed on the K VMFs to identify the target VMF corresponding to the first type of noise. Noise removal processing based on the target VMF may involve first determining the peak frequency corresponding to the maximum frequency domain amplitude of each VMF; then determining the frequency range based on the preset power line interference frequency and the expected error range; and finally filtering each VMF using a notch filter based on the peak frequency and the frequency range to achieve the removal of power line interference noise.
[0061] In some implementations, frequency analysis can be performed on the remaining K-1 VMFs after removing baseline drift noise to identify the target VMF corresponding to the first type of noise, and then noise removal can be performed based on the target VMF.
[0062] In some embodiments of this application, power line interference noise generally manifests as 50Hz fixed frequency noise and its harmonics. After performing VMD processing, power line interference noise and baseline drift noise are often confined to a sub-mode (VMF) of VMD.
[0063] Therefore, for power line interference noise, the frequency corresponding to the maximum amplitude of each VMF in the frequency domain can be calculated separately. The mode of each VMF near the preset power line interference frequency is identified, and a notch filter is used for filtering to remove power frequency power line interference noise.
[0064] The filtering using a notch filter can be expressed as:
[0065] ;
[0066] and ;
[0067] For the first The VMF in the frequency domain is the first The amplitude of each sampling point The total length of the frequency box, i.e. the total number of sampling points in the spectrum, can be determined based on the required frequency resolution and signal sampling frequency. For the denoised first VMF, Indicates a low quality factor with variation. The second-order notch filter, The symbol for convolution. To preset the power line interference frequency, To define the expected error range, in one example, you can set... and .
[0068] Figure 2 This is a flowchart illustrating a method for decomposing the target VMF after the first denoising process using RCMDE, as provided in an embodiment of this application. Figure 2 As shown, the method includes the following steps:
[0069] In step S201, the target variational mode function after the first denoising process is coarsened to obtain coarsened sequences at different time scales.
[0070] In step S202, the coarse-grained sequence is mapped to a normalized normal distribution sequence, and different scattering patterns are determined based on different target parameters.
[0071] The target parameters include at least the embedding dimension, the number of categories, and the time delay.
[0072] In step S203, the average mode probability of each scattering mode is calculated, and the fine composite multiscale scattering entropy value of the target variational mode function after the first denoising process is calculated based on the average mode probability.
[0073] In step S204, the signal components of the target variational mode function after the first denoising process are determined based on the fine composite multi-scale scattering entropy value.
[0074] The target VMF after the first denoising process is any VMF after the first denoising process.
[0075] In some embodiments of this application, when using RCMDE to decompose the target VMF after the first denoising process, the target VMF after the first denoising process can first be coarsened to obtain coarse-grained sequences at different time scales. Then, the coarse-grained sequences are mapped to normalized normal distribution sequences, and different scattering patterns are determined based on different target parameters. Next, the average mode probability of each scattering pattern is calculated, and the RCMDE value of the target VMF after the first denoising process is calculated based on the average mode probability. Finally, the signal components of the target VMF after the first denoising process are determined based on the RCMDE value.
[0076] In other words, in addition to the first type of noise, the original MMG signal may also contain a second type of noise. In some implementations, the second type of noise may include Gaussian white noise. Gaussian white noise has uniform power across all frequencies. After VMD, it is divided into VMFs with equal energy. Therefore, denoising for Gaussian white noise can be combined with wavelet thresholding denoising. An improved thresholding function is used to reduce Gaussian white noise in each VMF, and the noise and signal modes are extracted by combining RCMDE values. After soft thresholding, the signal is reconstructed to obtain the signal with noise artifacts removed.
[0077] The calculation of RCMDE values involves coarsening the various mode functions obtained from the decomposition to obtain coarse-grained sequences at different time scales, then mapping them to a normalized normal distribution sequence, obtaining different scattering patterns according to the embedding dimension, number of categories, and time delay, and calculating the RCMDE value by statistically analyzing the probabilities of different scattering patterns and then calculating the scattering entropy.
[0078] For any VMF, assume its length is... , No. The starting point at the th starting point Coarsening value Formulas can be used Calculated; where, The time scale is used to determine the size of the coarse-grained window, and corresponds to different starting points. Starting point It is the starting index used to generate different coarse-grained sequences. The value can be set according to actual needs. In one example, it can be set to... .
[0079] The average mode probability of different scattering patterns is ;in, For the first The probability of a scattering pattern at each starting point.
[0080] Furthermore, formulas can be used. The RCMDE values for each VMF were calculated; where, For the first The RCMDE value of each VMF, For embedded dimensions, For the number of categories, For time delay, It is the natural logarithm function. , and The specific value can be set according to actual needs. In one example, it can be set to... .
[0081] In some embodiments of this application, determining the signal components of the target VMF after the first denoising process based on the RCMDE value may include: determining the target VMF after the first denoising process as a noise component VMF in response to determining that the RCMDE value is greater than or equal to a first threshold; determining the target VMF after the first denoising process as a noise-dominant component VMF in response to determining that the RCMDE value is greater than or equal to a second threshold and less than the first threshold; determining the target VMF after the first denoising process as a signal-dominant component VMF in response to determining that the RCMDE value is greater than or equal to a third threshold and less than the second threshold; and determining the target VMF after the first denoising process as a signal component VMF in response to determining that the RCMDE value is less than the third threshold; wherein the first threshold is greater than the second threshold, and the second threshold is greater than the third threshold.
[0082] The specific values of the first, second, and third thresholds can be set according to actual needs. In one example, the first threshold can be set to 1.5, the second threshold to 1.0, and the third threshold to 0.5.
[0083] In some embodiments of this application, denoising the signal dominant component VMF and the noise dominant component VMF after local mean decomposition using wavelet soft thresholding may include: performing local mean decomposition on the signal dominant component VMF and the noise dominant component VMF to obtain M product functions; M is a positive integer; performing wavelet soft thresholding denoising on each product function to obtain a denoised product function; wherein the wavelet soft threshold corresponding to each product function is calculated independently; and synthesizing the denoised product functions to obtain the VMF after the second denoising process.
[0084] In some embodiments of this application, the wavelet soft threshold corresponding to the target product function can be determined in the following manner: ;in, For the first Wavelet soft thresholding corresponding to each objective product function As a compensation factor, For the first Noise standard deviation estimate of the objective product function For the first The length of each objective product function.
[0085] Furthermore, wavelet soft-thresholding denoising for each product function can include using the formula For the Wavelet soft thresholding denoising is performed on the product functions of the target functions; where... For the denoised first A target product function, For the first A target product function, It is a symbolic function.
[0086] In some embodiments of this application, signal reconstruction can be performed based on the VMF after the second denoising process. The processed modes (signal components after removing noise components, processed noise-dominant components, and signal-dominant components) are linearly superimposed to obtain the final denoised MMG signal.
[0087] Figure 3 This is a schematic flowchart illustrating another method for denoising myomagnetic signals based on improved variational mode decomposition and soft thresholding provided in this application. Figure 3 As shown, the device executing this method can first input a noisy MMG signal and initialize the VMD parameters. Then, the noisy signal is decomposed using VMD to obtain multiple VMFs.
[0088] Noise identification can be performed on these multiple VMFs. On the one hand, when removing power line interference (PLI) noise, it can be achieved by calculating the maximum frequency of the VMF frequency domain amplitude and combining it with a notch filter; on the other hand, it can be achieved by removing baseline drift (BW) noise through mode removal; and it can also remove white Gaussian noise (WGN).
[0089] When removing WGN, the RCMDE of each VMF can be calculated first. Then, a threshold judgment is performed on the RCMDE of each VMF to determine the noise component, the signal component, and the VMF corresponding to the noise or signal dominant component. Noise components are removed, while signal components are preserved. For both noise and signal dominant components, LMD decomposition is first performed, followed by wavelet soft thresholding for denoising. Finally, the VMFs that have undergone various denoising processes are recombined to obtain the denoised MMG signal.
[0090] Since BW noise and PLI noise are both narrowband noises, after VMD decomposition, these two types of noise are usually concentrated in two VMF modes. WGN noise, on the other hand, is broadband noise and is scattered across various modes. Therefore, we can first identify the modes containing BW noise and PLI noise after VMD decomposition, and remove or notch filter these modes. Then, we perform RCMDE analysis, LMD, and soft thresholding on the remaining modes. Finally, we perform linear synthesis on the processed modes. The synthesized signal in this way suppresses all three types of noise, preventing uncontrollable noise residue.
[0091] The denoising method provided in this application can be validated by comparing MMG simulation signals and actual acquired MMG signals. Taking the signal obtained from an actual MMG acquisition experiment as an example, the specific implementation process of this application's embodiment is illustrated below:
[0092] 1) Set VMD parameters: number of modes Punishment factor Tolerance error ;
[0093] 2) Perform VMD decomposition on the noisy MMG signal to obtain 12 VMFs;
[0094] 3) Identify and remove the first VMF (remove baseline drift noise);
[0095] 4) Identify and filter the VMF (power line interference noise) where the power frequency interference (50Hz) is located;
[0096] 5) Calculate the RCMDE of the remaining VMFs and perform modal classification;
[0097] 6) Perform LMD decomposition on the classified noise-dominant and signal-dominant components;
[0098] 7) Perform wavelet soft thresholding denoising on the decomposed PF;
[0099] 8) Reconstruct the PF after processing and superimpose it with the signal components to obtain the denoised MMG signal.
[0100] The acquired MMG signals were processed using the preprocessing framework designed in the above embodiment. The signals before and after denoising were evaluated by comparing them in the time and frequency domains, as well as by metrics such as signal-to-noise ratio (SNR), root mean square error (RMSE), and correlation coefficient. Whether the muscle activation time matches the actual experimental activation time, the residual noise in the resting state after preprocessing, and the clear distinction between the resting and active states after preprocessing can be used as direct observation targets. In the absence of the original clean signal, the SNR can be estimated using the amplitude values of the two states to evaluate the preprocessing effect.
[0101] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.
[0102] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.
[0103] Figure 4 This is a schematic diagram of a myomagnetic signal denoising device based on improved variational mode decomposition and soft thresholding provided in an embodiment of this application. Figure 4 As shown, the device includes:
[0104] The acquisition module 401 is configured to acquire the raw myoma magnetic resonance (MMG) signal; the raw MMG signal is a noisy signal.
[0105] The first decomposition module 402 is configured to perform variational mode decomposition (VMD) on the original MMG signal to obtain K variational mode functions (VMF); K is a positive integer.
[0106] The first denoising module 403 is configured to perform frequency analysis on K VMFs, identify the target VMF corresponding to the first type of noise, and perform noise removal processing based on the target VMF to obtain the VMF after the first denoising process.
[0107] The second decomposition module 404 is configured to decompose the VMF after the first denoising process using the fine composite multiscale scattering entropy (RCMDE) to determine the component types of each VMF; wherein the component types include noise components, noise-dominant components, signal components, and signal-dominant components.
[0108] The second denoising module 405 is configured to retain the signal component VMF and remove the noise component VMF, and to perform denoising processing on the signal dominant component VMF and the noise dominant component VMF after local mean decomposition using wavelet soft thresholding.
[0109] The reconstruction module 406 is configured to linearly superimpose the VMFs after the second denoising process to reconstruct the denoised MMG signal.
[0110] According to the technical solution provided in the embodiments of this application, multiple VMFs are obtained by performing VMD processing on the original MMG signal. Frequency analysis is then performed on these multiple VMFs to identify the target VMF corresponding to the first type of noise. The target VMF is then denoised to obtain the VMF after the first denoising process. Then, RCMDE is used to decompose the VMF after the first denoising process to determine the component type of each VMF. The signal component VMF is retained while the noise component VMF is removed. At the same time, the signal dominant component VMF and the noise dominant component VMF are denoised using wavelet soft thresholding after local mean decomposition. Finally, the VMFs after the second denoising process are linearly superimposed to reconstruct the denoised MMG signal. This method can effectively adapt to the non-stationary characteristics of MMG signals and accurately assess signal complexity, thereby improving the noise removal efficiency of MMG signals.
[0111] In some implementations, the first type of noise includes at least baseline drift noise; frequency analysis is performed on K VMFs to identify the target VMF corresponding to the first type of noise, and noise removal processing is performed based on the target VMF, including: locating the VMF where the baseline drift noise is located based on the characteristic of sorting the center frequencies of the K VMFs from low to high; the VMF where the baseline drift noise is located is the first VMF; the first VMF is discarded to achieve the removal of baseline drift noise.
[0112] In some implementations, the first type of noise includes at least power line interference noise; frequency analysis is performed on K VMFs to identify the target VMFs corresponding to the first type of noise, and noise removal processing is performed based on the target VMFs, including: determining the peak frequency corresponding to the maximum amplitude of each VMF in the frequency domain; determining the frequency range based on the preset power line interference frequency and the expected error range; and filtering each VMF using a notch filter based on the peak frequency and the frequency range to achieve the removal of power line interference noise.
[0113] In some implementations, RCMDE is used to decompose the target VMF after the first denoising process, including: coarsening the target VMF after the first denoising process to obtain coarse-grained sequences at different time scales; mapping the coarse-grained sequences to normalized normal distribution sequences, and determining different scattering patterns based on different target parameters; wherein the target parameters include at least the embedding dimension, the number of classes, and the time delay; calculating the average mode probability of each scattering pattern, and calculating the RCMDE of the target VMF after the first denoising process based on the average mode probability; and determining the signal components of the target VMF after the first denoising process based on the RCMDE value; wherein the target VMF after the first denoising process is any VMF after the first denoising process.
[0114] In some implementations, determining the signal components of the target VMF after the first denoising process based on the RCMDE value includes: determining the target VMF after the first denoising process as a noise component VMF in response to determining that the RCMDE value is greater than or equal to a first threshold; determining the target VMF after the first denoising process as a noise-dominant component VMF in response to determining that the RCMDE value is greater than or equal to a second threshold and less than the first threshold; determining the target VMF after the first denoising process as a signal-dominant component VMF in response to determining that the RCMDE value is greater than or equal to a third threshold and less than the second threshold; and determining the target VMF after the first denoising process as a signal component VMF in response to determining that the RCMDE value is less than the third threshold; wherein the first threshold is greater than the second threshold, and the second threshold is greater than the third threshold.
[0115] In some implementations, the signal-dominant component VMF and the noise-dominant component VMF are denoised using wavelet soft thresholding after local mean decomposition. This includes: performing local mean decomposition on the signal-dominant component VMF and the noise-dominant component VMF to obtain M product functions; M is a positive integer; performing wavelet soft thresholding denoising on each product function to obtain a denoised product function; wherein the wavelet soft threshold corresponding to each product function is calculated independently; and synthesizing the denoised product functions to obtain the VMF after the second denoising process.
[0116] In some implementations, the wavelet soft threshold corresponding to the target product function is determined in the following way: ;in, For the first Wavelet soft thresholding corresponding to each objective product function As a compensation factor, For the first Noise standard deviation estimate of the objective product function For the first The length of each target product function; wavelet soft thresholding denoising for each product function includes, using the formula For the Wavelet soft thresholding denoising is performed on the product functions of the target functions; where... For the denoised first A target product function, For the first A target product function, It is a symbolic function.
[0117] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0118] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 5 of this embodiment includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, it implements the steps in the various method embodiments described above. Alternatively, when the processor 501 executes the computer program 503, it implements the functions of each module / unit in the various device embodiments described above.
[0119] Electronic device 5 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 5 may include, but is not limited to, processor 501 and memory 502. Those skilled in the art will understand that... Figure 5This is merely an example of electronic device 5 and does not constitute a limitation on electronic device 5. It may include more or fewer components than shown, or different components.
[0120] The processor 501 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
[0121] The memory 502 can be an internal storage unit of the electronic device 5, such as a hard disk or RAM of the electronic device 5. The memory 502 can also be an external storage device of the electronic device 5, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc., equipped on the electronic device 5. The memory 502 can also include both internal and external storage units of the electronic device 5. The memory 502 is used to store computer programs and other programs and data required by the electronic device.
[0122] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0123] If an integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium may include: any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.
[0124] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. A method for denoising myoma signals based on improved variational mode decomposition and soft thresholding, characterized in that, include: Acquire raw myoma magnetic resonance (MMG) signals; The original MMG signal is a noisy signal; The original MMG signal is subjected to variational mode decomposition (VMD) to obtain K variational mode functions (VMFs); K is a positive integer. Frequency analysis is performed on the K VMFs to identify the target VMF corresponding to the first type of noise. Noise removal processing is then performed on the target VMF to obtain the VMF after the first denoising process. The VMF after the first denoising process is decomposed using the Refined Composite Multiscale Spread Entropy (RCMDE) to determine the component types of each VMF; wherein, the component types include noise components, noise-dominant components, signal components, and signal-dominant components. The signal component VMF is preserved while the noise component VMF is removed. After local mean decomposition, wavelet soft thresholding is used to denoise both the signal-dominant VMF and the noise-dominant VMF. The VMFs after the second denoising process are linearly superimposed to reconstruct the denoised MMG signal.
2. The method according to claim 1, characterized in that, The first type of noise includes at least baseline drift noise; Frequency analysis is performed on the K VMFs to identify the target VMF corresponding to the first type of noise. Noise removal processing is then performed based on the target VMF, including: Based on the characteristic of sorting the center frequencies of the K VMFs from low to high, the VMF containing the baseline drift noise is located; the VMF containing the baseline drift noise is the first VMF. The first VMF is discarded to remove baseline drift noise.
3. The method according to claim 1, characterized in that, The first type of noise includes at least power line interference noise; Frequency analysis is performed on the K VMFs to identify the target VMF corresponding to the first type of noise. Noise removal processing is then performed based on the target VMF, including: Determine the peak frequency corresponding to the maximum amplitude of each VMF in the frequency domain; The frequency range is determined based on the preset power line interference frequency and the expected error range; Based on the peak frequency and the frequency range, a notch filter is used to filter each VMF to remove power line interference noise.
4. The method according to claim 1, characterized in that, The target VMF after the first denoising process is decomposed using RCMDE, including: The target VMF after the first denoising process is coarse-grained to obtain coarse-grained sequences at different time scales; The coarse-grained sequence is mapped to a normalized normal distribution sequence, and different scattering patterns are determined based on different target parameters; wherein, the target parameters include at least the embedding dimension, the number of categories, and the time delay; The average mode probability of each scattering pattern is calculated, and the RCMDE value of the target VMF after the first denoising process is calculated based on the average mode probability. The signal components of the target VMF after the first denoising process are determined based on the RCMDE value. Wherein, the target VMF after the first denoising process is any VMF after the first denoising process.
5. The method according to claim 4, characterized in that, Determining the signal components of the target VMF after the first denoising process based on the RCMDE value includes: In response to determining that the RCMDE value is greater than or equal to the first threshold, the target VMF after the first denoising process is determined to be a noise component VMF; In response to determining that the RCMDE value is greater than or equal to the second threshold and less than the first threshold, the target VMF after the first denoising process is determined to be the noise-dominant component VMF; In response to determining that the RCMDE value is greater than or equal to the third threshold and less than the second threshold, the target VMF after the first denoising process is determined to be the dominant component VMF of the signal. In response to determining that the RCMDE value is less than the third threshold, the target VMF after the first denoising process is determined to be the signal component VMF; Wherein, the first threshold is greater than the second threshold, and the second threshold is greater than the third threshold.
6. The method according to claim 1, characterized in that, After local mean decomposition, wavelet soft thresholding is used to denoise both the signal-dominant VMF and the noise-dominant VMF, including: Local mean decomposition is performed on the signal-dominant component VMF and the noise-dominant component VMF to obtain M product functions; M is a positive integer; Each product function is subjected to wavelet soft thresholding denoising to obtain the denoised product function; the wavelet soft threshold for each product function is calculated independently. The VMF after the second denoising process is obtained by synthesizing the product function after denoising.
7. The method according to claim 6, characterized in that, The wavelet soft threshold corresponding to the target product function is determined as follows: ； in, For the first Wavelet soft thresholding corresponding to each objective product function As a compensation factor, For the first Noise standard deviation estimate of the objective product function For the first The length of each objective product function; Wavelet soft-thresholding denoising for each product function includes using the formula For the Wavelet soft thresholding denoising is performed on the product functions of the target functions; where... For the denoised first A target product function, For the first A target product function, It is a symbolic function.
8. A myoma signal denoising device based on improved variational mode decomposition and soft thresholding, characterized in that, include: The acquisition module is configured to acquire the raw myoma magnetic resonance (MMG) signal; the raw MMG signal is a noisy signal. The first decomposition module is configured to perform variational mode decomposition (VMD) on the original MMG signal to obtain K variational mode functions (VMF); where K is a positive integer. The first denoising module is configured to perform frequency analysis on the K VMFs, identify the target VMF corresponding to the first type of noise, and perform noise removal processing based on the target VMF to obtain the VMF after the first denoising process. The second decomposition module is configured to decompose the VMF after the first denoising process using the Refined Composite Multiscale Spread Entropy (RCMDE) to determine the component types of each VMF; wherein the component types include noise components, noise-dominant components, signal components, and signal-dominant components. The second denoising module is configured to retain the signal component VMF and remove the noise component VMF, and to perform denoising processing on the signal dominant component VMF and the noise dominant component VMF after local mean decomposition using wavelet soft thresholding. The reconstruction module is configured to linearly superimpose the VMFs after the second denoising process to reconstruct the denoised MMG signal.
9. 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, it implements the steps of the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.