A method and system for removing electro-physiological signal gradient artifacts in a magnetic resonance environment

Abstract

The application provides a method and system for removing electro-physiological signal gradient artifacts in a magnetic resonance environment, and relates to the technical field of magnetic resonance image processing. The method comprises: collecting an electro-physiological signal of a magnetic resonance scan and a scan synchronization signal; aligning and windowing the electro-physiological signal based on the scan synchronization signal to obtain a windowed sequence; constructing an artifact template based on the windowed sequence; combining the windowed sequence and the artifact template to calculate a first residual; performing principal component analysis on the artifact template and the first residual to obtain a second residual; performing logarithmic domain conversion and polarity extraction operations on the second residual to obtain a logarithmic domain sequence and a polarity symbol layer; inputting the logarithmic domain sequence into a deep learning denoising network to output a de-artifact logarithmic domain sequence; and performing inverse logarithmic transformation and polarity recovery operations on the de-artifact logarithmic domain sequence and the polarity symbol layer, respectively, to determine a de-artifact signal.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance imaging technology, and in particular to a method and system for removing gradient artifacts of electrophysiological signals in a magnetic resonance environment. Background Technology

[0002] Simultaneous acquisition of electrophysiological signals such as electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG) during magnetic resonance imaging (MRI) scans is an important research direction in the field of electrophysiological signal processing and MRI-compatible monitoring technology. Real-time electrophysiological monitoring and analysis under high-field MRI and rapid sequences have become important development needs in this field. Rapid switching of the MRI gradient magnetic field is the core element for achieving imaging. The gradient artifacts induced by the MRI overlap with the frequency band of the target signal, and the synchronous acquisition system must balance imaging requirements with the effective extraction of electrophysiological signals.

[0003] In the current field of MRI electrophysiological signal acquisition, at the hardware level, shielding structures, amplifiers placed inside the MRI chamber, and fiber optic isolation transmission are employed to reduce additional interference in the acquisition link, providing a basic guarantee for signal acquisition. At the software level, average artifact template subtraction and optimal basis set methods have become common means of gradient artifact suppression. Some studies have introduced deep learning models to achieve end-to-end mapping of artifacts. These various technologies have laid a diverse technical foundation for the acquisition and processing of MRI-synchronized electrophysiological signals.

[0004] However, existing template-based methods are susceptible to artifact morphology changes, and deep learning methods are prone to error amplification when the gradient artifact amplitude is much larger than the real signal. Furthermore, some methods require additional hardware structures, raising the adaptation threshold for venues and equipment. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method for removing gradient artifacts of electrophysiological signals in a magnetic resonance environment. This method can solve the problems of existing template-based methods being easily affected by changes in artifact morphology, deep learning methods being prone to error amplification when the gradient artifact amplitude is much larger than the real signal, and some methods requiring additional hardware structures, which raises the technical threshold for site and equipment adaptation.

[0006] A first aspect of this invention provides a method for removing gradient artifacts of electrophysiological signals in a magnetic resonance imaging (MRI) environment, comprising:

[0007] S1: Acquire electrophysiological signals and scan synchronization signals from magnetic resonance imaging (MRI) scans;

[0008] S2: Based on the scanning synchronization signal, the electrophysiological signals are aligned and windowed to obtain a windowed sequence;

[0009] S3: Construct artifact templates based on windowed sequences;

[0010] S4: Calculate the first residual by combining the windowed sequence and the artifact template;

[0011] S5: Perform principal component analysis on the artifact template and the first residual to obtain the second residual;

[0012] S6: Perform logarithmic field transformation and polarity extraction operations on the second residual to obtain the logarithmic field sequence and polarity sign layer;

[0013] S7: Input the logarithmic domain sequence into a deep learning denoising network and output a de-artifacted logarithmic domain sequence;

[0014] S8: Based on the artifact-removed logarithmic domain sequence and polar sign layer, perform anti-logarithmic transformation and polarity recovery operations respectively to determine the artifact-removed signal.

[0015] A second aspect of the present invention provides a system for removing gradient artifacts of electrophysiological signals in a magnetic resonance imaging environment, comprising: a processor and a memory;

[0016] The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the electrophysiological signal gradient artifact removal method in a magnetic resonance environment as described in the first aspect.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0018] In this embodiment of the invention, to address the shortcomings of existing template-based methods being susceptible to artifact morphology changes, deep learning methods suffering from error amplification, and some methods requiring additional hardware, this method provides targeted solutions through algorithmic processing. Principal component analysis is performed on the artifact template and the first residual to obtain the second residual. This residual is then fitted to subtract artifact morphology changes, effectively improving the adaptability of template-based methods. The second residual undergoes logarithmic domain transformation and polarity extraction sequentially. After processing by a deep learning denoising network, it is further subjected to antilogarithmic transformation and polarity restoration, mitigating the error amplification problem of deep learning while ensuring accurate signal polarity. This method achieves artifact removal entirely through the algorithmic processing steps of electrophysiological signals, requiring no additional hardware structure, lowering the application adaptation threshold, and enabling effective extraction of electrophysiological signals in a magnetic resonance imaging (MRI) environment. Attached Figure Description

[0019] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0020] Figure 1 This is a schematic flowchart of a method for removing gradient artifacts of electrophysiological signals in a magnetic resonance environment, provided by an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of an electrophysiological signal gradient artifact removal system under magnetic resonance environment provided in an embodiment of the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] The method for removing gradient artifacts of electrophysiological signals in a magnetic resonance environment provided by the present invention will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0024] Reference manual attached Figure 1 The diagram illustrates a flowchart of a method for removing gradient artifacts of electrophysiological signals in a magnetic resonance environment, as provided in an embodiment of the present invention.

[0025] This invention provides a method for removing gradient artifacts of electrophysiological signals under magnetic resonance imaging, which may include the following steps:

[0026] S1: Acquire electrophysiological signals and scan synchronization signals from magnetic resonance imaging (MRI) scans.

[0027] Optionally, the electrophysiological signals specifically include electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, and electromyogram (EMG) signals.

[0028] The scanning synchronization signal is specifically the slice timing signal or trigger signal of the magnetic resonance imaging sequence.

[0029] Among them, electrophysiological signals refer to bioelectrical signals such as electroencephalogram (EEG), electrocardiogram (ECG), or electromyogram (EMG) acquired synchronously during magnetic resonance imaging (MRI) scanning, while scanning synchronization signals refer to the slice timing signals or trigger signals output by the MRI sequence, which are used to identify the start time of the gradient switching cycle.

[0030] Specifically, the core is to simultaneously acquire two types of key signals: one is the electrophysiological signal that reflects human physiological activities, and the other is the scanning synchronization signal used for subsequent timing alignment. Together, they constitute the basic input for subsequent artifact removal processing.

[0031] S2: Based on the scanning synchronization signal, the electrophysiological signal is aligned and windowed to obtain a windowed sequence.

[0032] Among them, windowed sequence refers to the isochronous signal segments obtained by segmenting continuous electrophysiological signals with the scanning synchronization signal as the boundary, and each segment covers a complete gradient switching cycle.

[0033] Specifically, the windowed sequence is defined by the following formula:

[0034]

[0035] in, c Indicates the channel number. k Indicates the window number. n Indicates the sampling point number within the window. n 0( k ) indicates the first k The starting position of each window in the original continuous electrophysiological signal. N w Indicates the window length. x c,k [ n ] indicates the windowed sequence, the first... c The first channel, the first k The first window n The signal value at each sampling point.

[0036] It should be noted that the alignment and windowing process also includes estimating the optimal time shift between adjacent windows by finding peaks through cross-correlation, and using subsampling interpolation to achieve subsampling-level alignment to ensure the consistency of artifact morphology within the window.

[0037] Specifically, the aligned windowed sequence can be approximated as:

[0038]

[0039] in, This represents the aligned windowed sequence. Indicates the first k The optimal time shift for each window.

[0040] Specifically, the processing strictly relies on the timing identification function of the scanning synchronization signal. The continuous electrophysiological signal is segmented using this signal as the boundary. In the final windowed sequence, each signal segment corresponds to a complete gradient switching cycle, ensuring the periodic consistency of subsequent artifact template construction.

[0041] S3: Construct artifact templates based on windowed sequences.

[0042] Among them, the artifact template refers to the estimated waveform that represents the dominant morphology of gradient artifacts, generated by statistically analyzing the commonalities and periodic structures of multiple windowed sequences.

[0043] Specifically, the core objective of S3 is to construct an artifact template that can characterize the dominant pattern of gradient artifacts. The template is built on a windowed sequence. By statistically analyzing the common features and periodic structure of multiple windowed sequences, an estimated waveform for subsequent artifact subtraction is finally generated.

[0044] In one possible implementation, S3 specifically includes sub-steps S301 to S304:

[0045] S301: Determine the set of neighboring windows of the current window.

[0046] The neighborhood window set refers to a continuous group of windows centered on the current processing window and consisting of several windows taken before and after it on the timeline.

[0047] Specifically, by selecting adjacent windows of the current processing window, and taking the current window as the center, several windows are selected before and after it on the time axis. These windows together form a neighborhood window set, providing basic samples for subsequent similarity calculation.

[0048] S302: Calculate the similarity between windows in the neighborhood window set based on the windowed sequence.

[0049] Among them, similarity refers to a quantitative indicator that measures the consistency of waveform morphology between different windowed sequences.

[0050] Furthermore, the similarity calculation relies on the waveform features of the windowed sequence. By quantifying the consistency of waveform morphology among windows in the neighborhood window set, the similarity result of each window with other windows is obtained, providing a basis for judgment in subsequent window processing.

[0051] S303: Based on similarity, perform weighted fusion processing on the neighborhood window set to obtain the processed window set.

[0052] The processing refers to removing or downsizing low-similarity windows based on a similarity threshold to reduce the contamination of the template by factors such as head movement and breathing.

[0053] Specifically, the judgment criterion for the processing operation is the similarity calculation result obtained by S302. Based on the preset similarity threshold, windows with low similarity in the neighborhood window set are removed or downweighted, thereby reducing the contamination of the artifact template by irrelevant factors such as head movement and breathing, and ensuring the accuracy of the template.

[0054] It should be noted that those skilled in the art can set the preset similarity threshold according to actual needs, and this invention does not limit it.

[0055] S304: Average the processed window set to construct an artifact template.

[0056] It should be noted that the number of windows K in the neighborhood window set can be between 5 and 50. The similarity threshold can be set based on normalized cross-correlation or cosine similarity. When the similarity is lower than the similarity threshold, the window is removed or given a smaller weight.

[0057] Specifically, the artifact template is constructed using an averaging method, with the following formula:

[0058]

[0059] in, c Indicates the channel number. k Indicates the window number. n Indicates the sampling point number within the window. Indicates the first c The first channel, the first k The first artifact template in the window n The signal value at each sampling point Ω K Represents the processed set of neighborhood windows. i Represents the set of neighborhood windows Ω K Window number in w k,i Indicates the first k The set of windows and neighboring windows i The weight of each window, This represents the aligned windowed sequence.

[0060] S4: Calculate the first residual by combining the windowed sequence and the artifact template.

[0061] The first residual refers to the signal component retained after subtracting the artifact template from the windowed sequence of the current window, including the residual artifacts that were not subtracted and the real electrophysiological signal.

[0062] Specifically, the first residual is obtained through the following formula:

[0063]

[0064] in, c Indicates the channel number. k Indicates the window number. n Indicates the sampling point number within the window. Indicates the first c The first channel, the first k The first residual of the window n The signal value at each sampling point This represents the aligned windowed sequence. Indicates the first c The first channel, the first k The first artifact template in the window n The signal value at each sampling point.

[0065] Specifically, the first residual is obtained by subtracting the windowed sequence of the current window from the artifact template constructed by S3. The signal component retained after the operation is the first residual, which includes the residual artifacts that were not subtracted by the artifact template and the real electrophysiological signals that need to be extracted.

[0066] S5: Perform principal component analysis on the artifact template and the first residual to obtain the second residual.

[0067] Principal component analysis (PCA) is a statistical method that extracts the main change direction from the artifact template and the first residual through linear transformation. The second residual is the signal obtained by fitting the basis vectors extracted by PCA to the artifact changes in the current window and then subtracting the fitted components from the first residual.

[0068] Furthermore, firstly, using the statistical method of principal component analysis, the basis vectors of the main change directions in the artifact template and the first residual are extracted. Then, the artifact changes in the current window are fitted using these basis vectors to obtain the artifact fitting components. Finally, the first residual is subtracted from the fitting components to obtain the second residual, thereby achieving further suppression of residual artifacts.

[0069] Furthermore, the number of basis vectors M extracted by principal component analysis is 2 to 20, which can be selected according to the cumulative variance contribution rate threshold of 90% to 99%, or determined according to the preset upper limit value, in order to avoid overfitting.

[0070] It should be noted that those skilled in the art can set the variance contribution rate threshold and the preset upper limit value according to actual needs, and this invention does not limit them.

[0071] Specifically, the optimal basis set is represented as:

[0072]

[0073] in, B Denotes the optimal basis set matrix. b 1 represents the optimal basis set matrix B The first basis vector in, b 2 represents the optimal basis set matrix. B The second basis vector in, b M Describing the optimal basis set matrix B The first in M basis vectors M Describing the optimal basis set matrixB The number of basis vectors in the middle.

[0074] The basis coefficients are obtained through least squares fitting:

[0075]

[0076] in, c Indicates the channel number. k Indicates the window number. β k Indicates the first c The first channel, the first k The basis coefficient vector within a window Indicates the first c The first channel, the first k The first residual vector of each window β This represents the vector of basic coefficients to be solved. This represents the square of the L2 norm.

[0077] The second residual is obtained through the following formula:

[0078]

[0079] in, Indicates the first c The first channel, the first k The second residual of the window n The signal value at each sampling point Indicates the first c The first channel, the first k The first residual of the window n The signal value at each sampling point Bβ k Indicates the first c The first channel, the first k The artifact fitting component vector of each window, Represents the component vector of the artifact fitting Bβ k The Middle n The signal value at each sampling point.

[0080] Optionally, after S5 and before S6, the following are also included:

[0081] Perform a multi-channel Wiener filter operation on the second residual to obtain the third residual.

[0082] Multi-channel Wiener filtering refers to a method of spatial filtering of signals using the covariance structure between multiple channels and under the minimum mean square error criterion. The third residual is the output signal after suppressing cross-channel correlation residual artifacts through multi-channel Wiener filtering.

[0083] Specifically, multi-channel Wiener filtering further suppresses residual artifacts related to cross-channel correlation. It relies on the covariance structure between multi-channel signals and performs spatial filtering on the second residual under the minimum mean square error criterion. The signal output after filtering is the third residual, which further improves the signal purity.

[0084] Optionally, performing a multi-channel Wiener filtering operation on the second residual specifically includes:

[0085] Identify the artifact-dominant and low-artifact periods in the second residual.

[0086] The artifact-dominant period refers to the time interval where the gradient artifact energy is significantly higher than the electrophysiological signal. The low-artifact period refers to the time interval where the gradient artifact energy is relatively weak or within the scanning interval.

[0087] Specifically, the time interval of the second residual is divided into two periods based on the energy ratio of gradient artifacts and electrophysiological signals: an artifact-dominated period and a low-artifact period. This division of the two periods provides a basis for the subsequent construction of the covariance matrix.

[0088] Based on the artifact-dominant time period and the second residual, a noise covariance matrix is ​​constructed.

[0089] The noise covariance matrix is ​​used to describe the cross-channel correlation structure of residual artifacts in multiple channels during the artifact-dominant period.

[0090] Furthermore, the construction of the noise covariance matrix relies on the artifact-dominant period of the second residual. Since the gradient artifact energy dominates during this period, the second residual of this period can be regarded as a noise signal. By calculating the covariance of the multi-channel signals during this period, the noise covariance matrix used to describe the correlation structure of cross-channel residual artifacts is obtained.

[0091] The signal covariance matrix is ​​constructed based on the low artifact period and the second residual.

[0092] The signal covariance matrix is ​​used to describe the cross-channel spatial distribution characteristics of real electrophysiological signals during periods with low artifacts.

[0093] Specifically, the construction of the signal covariance matrix relies on the low artifact period of the second residual, during which the energy of the real electrophysiological signal is dominant. By calculating the covariance of the second residual of multiple channels during this period, a signal covariance matrix is ​​obtained to describe the cross-channel spatial distribution characteristics of the real electrophysiological signal.

[0094] The filter coefficients are calculated by combining the noise covariance matrix and the signal covariance matrix.

[0095] Here, the filter coefficients refer to the linear transformation matrix of the multi-channel Wiener filter.

[0096] The second residual is filtered using the filtering coefficients to obtain the third residual.

[0097] Specifically, the second residual is used to construct a channel vector at each time step:

[0098]

[0099] in, Indicates the first k The first window, the n The second residual of each sampling point Indicates the first channel, the... k The first window, the... n The second residual of each sampling point Indicates the second channel, the... k The first window, the... n The second residual of each sampling point T This indicates the transpose operation.

[0100] The filter coefficients are determined using the standard formula for multi-channel Wiener filtering, specifically:

[0101]

[0102] in, W This represents the multi-channel Wiener filter coefficient matrix. R s Represents the signal covariance matrix. R v Let represent the noise covariance matrix, ( ) -1 This represents the inverse operation of a matrix.

[0103] The third residual is output after further filtering:

[0104]

[0105] in, Indicates the first k The first window, the n The third residual of each sampling point.

[0106] Specifically, the filter coefficients obtained above are applied to the second residual, and spatial filtering is performed on the second residual. By filtering, residual artifacts related to cross channels are suppressed, and finally the third residual is obtained.

[0107] S6: Perform logarithmic field transformation and polarity extraction operations on the second residual to obtain the logarithmic field sequence and polarity sign layer.

[0108] Among them, the logarithmic domain sequence refers to the numerical sequence obtained by logarithmically compressing the signal amplitude. The polarity sign layer refers to the binary label sequence used to record the positive and negative polarity information of the signal.

[0109] Specifically, the second residual (or the third residual if multi-channel Wiener filtering is performed) is processed in two parallel steps: logarithmic domain transformation and polarity extraction. These two steps yield the logarithmic domain sequence and the polarity sign layer, respectively, providing a foundation for subsequent deep learning denoising and signal recovery.

[0110] Optionally, the logarithmic field transformation of the second residual in S6 specifically includes:

[0111] Perform an absolute value transformation on the second residual to obtain the amplitude sequence.

[0112] Among them, the amplitude sequence refers to the non-negative sequence that retains only the numerical value after discarding the sign of the original signal.

[0113] Specifically, an absolute value transformation is performed on the second residual, discarding the sign of each sequence point in the second residual and retaining only its numerical value, ultimately obtaining a non-negative amplitude sequence, which prepares for subsequent logarithmic compression transformation.

[0114] Perform a logarithmic compression transformation on the amplitude sequence to obtain a logarithmic domain sequence.

[0115] Optionally, the polarity extraction operation on the second residual in S6 specifically includes:

[0116] Based on the polarity characteristics of the second residual, the polarity of multiple sequence points in the second residual is determined.

[0117] Among them, polarity characteristics refer to the positive and negative attributes of the signal waveform at each sampling point on the time axis.

[0118] Specifically, based on the polarity characteristics of the second residual, the positive or negative attribute of each sequence point in the second residual is determined, clarifying whether each sequence point is positive or negative, thus completing the determination of the polarity of the sequence points.

[0119] A polarity symbol layer is constructed based on the polarity of each sequence point.

[0120] Among them, the polarity symbol layer is a binary sequence of the same length as the signal sequence, with positive values ​​denoted as 1 and negative values ​​denoted as 0.

[0121] It should be noted that the logarithmic compression transformation is implemented using the following formula:

[0122]

[0123] in, c Indicates the channel number. k Indicates the window number. n Indicates the sampling point number within the window. a c,k [ n ] indicates the first c The first channel, the firstk The first window, the n Amplitude sequence of sampling points, Indicates the first c The first channel, the first k The third residual of the window n The signal value at each sampling point This represents absolute value operations.

[0124] Specifically, the logarithmic compression transformation uses the standard formula, which is as follows:

[0125]

[0126] in, z c,k [ n ] indicates the first c The first channel, the first k The first window, the n The logarithmic field sequence of n sampling points, where log() represents the logarithmic operation. α This represents the amplitude compression factor.

[0127] Specifically, the polarity symbol layer is constructed using a binary formula, as follows:

[0128]

[0129] in, s c,k [ n ] indicates the first c The first channel, the first k The first window, the n The polarity symbol layer of each sampling point.

[0130] S7: Input the logarithmic domain sequence into the deep learning denoising network and output the artifact-free logarithmic domain sequence.

[0131] Among them, deep learning denoising networks refer to neural network models that learn the mapping relationship from noisy inputs to clean outputs through a large amount of training data. Artifact-free logarithmic domain sequences refer to the clean amplitude compressed estimate of the network's output after suppressing residual artifacts in the logarithmic domain.

[0132] Specifically, the obtained logarithmic domain sequence is input into a deep learning denoising network. This network suppresses residual artifacts in the logarithmic domain sequence by learning the mapping relationship between noisy input and clean output through pre-training, and finally outputs an artifact-free logarithmic domain sequence.

[0133] Optionally, the deep learning denoising network is specifically a two-layer denoising autoencoder network, wherein the two-layer denoising autoencoder network includes a first denoising autoencoder and a second denoising autoencoder.

[0134] Among them, the dual-layer denoising autoencoder network refers to a deep neural network composed of two cascaded denoising autoencoders, with the output of the first denoising autoencoder serving as the input of the second denoising autoencoder.

[0135] Specifically, the deep learning denoising network adopts a two-layer cascaded structure, consisting of a first denoising autoencoder and a second denoising autoencoder. The two denoising autoencoders are connected in a cascaded manner, that is, the output signal of the first denoising autoencoder is directly used as the input signal of the second denoising autoencoder, thereby realizing hierarchical denoising.

[0136] In one possible implementation, S7 specifically includes sub-steps S701 and S702:

[0137] S701: Input the logarithmic domain sequence into a two-layer denoising autoencoder network.

[0138] Specifically, the obtained logarithmic domain sequence is directly input into a two-layer denoising autoencoder network as the network's input signal to initiate the network's denoising process.

[0139] S702: By using a dual-layer denoising autoencoder network, dual encoding and dual decoding operations are performed on the logarithmic domain sequence to output an artifact-free logarithmic domain sequence.

[0140] In this context, dual encoding refers to the network performing feature compression and abstract extraction on the input in the first and second layers, respectively. Dual decoding refers to the network recovering the signal structure from the compressed features in the first and second layers, respectively.

[0141] Furthermore, the dual-layer denoising autoencoder network performs dual encoding and dual decoding operations on the input logarithmic domain sequence. The first layer of encoding performs preliminary feature compression and abstraction extraction on the input, and the first layer of decoding initially recovers the signal structure from the compressed features. The second layer of encoding further compresses and extracts features from the result of the first layer of decoding, and the second layer of decoding further recovers the signal structure from the features of the second layer of encoding. Through hierarchical processing, the final output is a logarithmic domain sequence without artifacts.

[0142] S8: Based on the artifact-removed logarithmic domain sequence and polar sign layer, perform anti-logarithmic transformation and polarity recovery operations respectively to determine the artifact-removed signal.

[0143] Among them, the artifact removal signal is the final output electrophysiological signal in which gradient artifacts have been effectively suppressed.

[0144] Specifically, the antilogarithmic transform uses the following formula to restore the logarithmic domain magnitude estimate to the original magnitude scale:

[0145]

[0146] in, Indicates the first cThe first channel, the first k The first window, the n The magnitude estimation sequence of sampling points, exp() represents the exponential operation. Indicates the first c The first channel, the first k The first window, the n The artifact-free logarithmic domain sequence of each sampling point α This represents the amplitude compression factor.

[0147] Polarity recovery uses the following formula to restore the amplitude sequence into a complete signal with positive and negative waveforms based on the polarity sign layer:

[0148]

[0149] in, y c,k [ n ] indicates the first c The first channel, the first k The first window, the n The artifact-free signal of each sampling point s c,k [ n ] indicates the first c The first channel, the first k The first window, the n The polarity symbol layer of each sampling point.

[0150] Specifically, the output logarithmic domain sequence with artifact removal undergoes two parallel inverse processes. One is an antilogarithmic transformation, which restores the logarithmic domain sequence to its original amplitude scale. The other is polarity recovery, which restores the original positive and negative polarities of the signal based on the obtained polarity sign layer. After these two processes are completed, the final output is an artifact-removed signal with gradient artifacts effectively suppressed, thus completing the entire artifact removal process.

[0151] Reference manual attached Figure 2 The diagram shows a schematic of the structure of an electrophysiological signal gradient artifact removal system under magnetic resonance environment provided by an embodiment of the present invention.

[0152] This invention provides a system 20 for removing gradient artifacts of electrophysiological signals in a magnetic resonance imaging environment, comprising: a processor 201 and a memory 202;

[0153] The memory 202 stores programs or instructions that can run on the processor 201. When the program or instructions are executed by the processor 201, they implement the steps of the above-described method for removing electrophysiological signal gradient artifacts in a magnetic resonance environment and achieve the same technical effect. To avoid repetition, the present invention will not elaborate further.

[0154] It should be understood that the processor 201 in this embodiment of the invention may be a central processing unit (CPU), or it may be 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. The general-purpose processor may be a microprocessor or any conventional processor.

[0155] It should also be understood that the memory 202 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM).

[0156] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0157] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers 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 the present invention.

[0158] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0159] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0160] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0161] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0162] In addition, the functional units in the various embodiments of the present invention 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.

[0163] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0164] This invention provides a readable storage medium comprising: storing a program or instructions on the readable storage medium, wherein when the program or instructions are executed by a processor, the program or instructions implement the steps of the above-described method for removing electrophysiological signal gradient artifacts in a magnetic resonance environment, and can achieve the same technical effect. To avoid repetition, this invention will not elaborate further.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention 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; and these 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 the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A method for removing gradient artifacts of electrophysiological signals under magnetic resonance imaging, characterized in that, include: S1: Acquire electrophysiological signals and scan synchronization signals from magnetic resonance imaging (MRI) scans; S2: Based on the scanning synchronization signal, the electrophysiological signal is aligned and windowed to obtain a windowed sequence; S3: Construct an artifact template based on the windowed sequence; S3 specifically includes: S301: Determine the set of neighboring windows of the current window; S302: Based on the windowing sequence, calculate the similarity between each window in the neighborhood window set; S303: Based on the similarity, perform weighted fusion processing on the neighborhood window set to obtain the processed window set; S304: The processed window set is averaged to construct the artifact template; S4: Calculate the first residual by combining the windowed sequence and the artifact template; The first residual refers to the signal component retained after subtracting the artifact template from the windowed sequence of the current window, including the residual artifacts that were not subtracted and the real electrophysiological signal. S5: Perform principal component analysis on the artifact template and the first residual to obtain the second residual; The second residual refers to the signal obtained by subtracting the fitted component from the first residual after fitting the basis vector extracted by principal component analysis to the current window artifact change; S6: Perform logarithmic field transformation and polarity extraction operations on the second residual to obtain a logarithmic field sequence and a polarity sign layer; S7: Input the logarithmic domain sequence into a deep learning denoising network and output a de-artifacted logarithmic domain sequence; S8: Based on the artifact removal logarithmic domain sequence and the polarity sign layer, perform antilogarithmic transformation and polarity recovery operations respectively to determine the artifact removal signal.

2. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 1, characterized in that, The electrophysiological signals specifically include electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, and electromyogram (EMG) signals; The scanning synchronization signal is specifically a slice timing signal or trigger signal for the magnetic resonance imaging sequence.

3. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 1, characterized in that, The process includes the following steps after S5 and before S6: Perform a multi-channel Wiener filter operation on the second residual to obtain the third residual.

4. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 3, characterized in that, Performing a multi-channel Wiener filter operation on the second residual specifically includes: Determine the artifact-dominant and low-artifact periods in the second residual; Based on the artifact-dominant time period and the second residual, a noise covariance matrix is ​​constructed; Based on the low artifact period and the second residual, a signal covariance matrix is ​​constructed; The filter coefficients are calculated by combining the noise covariance matrix and the signal covariance matrix. The second residual is filtered using the filtering coefficients to obtain the third residual.

5. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 1, characterized in that, The logarithmic domain transformation of the second residual in S6 specifically includes: Perform an absolute value transformation on the second residual to obtain the amplitude sequence; The magnitude sequence is subjected to a logarithmic compression transformation to obtain the logarithmic domain sequence.

6. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 1, characterized in that, The polarity extraction operation of the second residual in S6 specifically includes: Based on the polarity characteristics of the second residual, determine the polarity of multiple sequence points in the second residual; The polarity symbol layer is constructed based on the polarity of each sequence point.

7. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 1, characterized in that, The deep learning denoising network is specifically a two-layer denoising autoencoder network, wherein the two-layer denoising autoencoder network includes a first denoising autoencoder and a second denoising autoencoder.

8. The method for removing electrophysiological signal gradient artifacts under magnetic resonance imaging according to claim 7, characterized in that, Specifically, S7 includes: S701: Input the logarithmic domain sequence into the two-layer denoising autoencoder network; S702: The dual-layer denoising autoencoder network performs dual encoding and dual decoding operations on the logarithmic domain sequence to output the artifact-free logarithmic domain sequence.

9. A system for removing electrophysiological signal gradient artifacts under magnetic resonance imaging, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the electrophysiological signal gradient artifact removal method in a magnetic resonance environment as described in any one of claims 1 to 8.

Images