A multivariate neural decoding method based on OPM-MEG and RSA

Abstract

The present invention discloses a multivariate neural decoding method based on OPM-MEG and RSA, which belongs to the field of biomedical engineering technology. The method comprises: obtaining resting-state structural MRI data and OPM-MEG scan images for joint registration; playing auditory stimulation and completing OPM-MEG data acquisition, preprocessing the data, and obtaining segmented data; averaging different trials of the same auditory stimulation at the segmented data level and constructing a neural response pattern vector, calculating the Pearson correlation of all possible pairwise stimulus pairs at each time point, obtaining a representation similarity matrix and an average result of the neural response pattern similarity; determining the time points with differences at each time point; estimating the source activation time series of each subject, and determining the activation area using a T-test. The present invention can reveal the decoding representation process of the brain by calculating the distance of the neural response pattern in the representation space, thereby more comprehensively understanding the response pattern of the brain under specific tasks or conditions.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical engineering technology, and in particular relates to a multivariate neural decoding method based on OPM-MEG and RSA. Background Art

[0002] In the field of neuroscience, the brain converts external stimuli into intentions through decoding. By decoding neural activity, we can understand how the brain processes information and generates thoughts and behaviors. This is of great significance to many fields, including neuroscience, psychology, medicine, engineering, and computer science.

[0003] In recent years, a variety of measurement modes for decoding brain neural activity have been developed. In non-invasive brain function measurement, optically pumped magnetometer-based magnetoencephalography (OPM-MEG) can directly measure the magnetic activity of neurons and capture the dynamic changes of the brain at the millisecond level. It has become an important brain function imaging technology with broad application prospects.

[0004] In addition to measurement patterns, methods for decoding neural activity have also made significant progress, mainly including two categories: univariate analysis and multivariate analysis. Among them, multivariate analysis can simultaneously consider the activity patterns of multiple brain regions, integrate the overall information of the brain, and more comprehensively understand the activity patterns of the brain under specific tasks or conditions.

[0005] Representation Similarity Analysis (RSA), as a method of multivariate analysis, provides a powerful explanatory framework through the core concept of "representation space", in which the neural response to each stimulus is represented as a high-dimensional pattern vector.

[0006] RSA has traditionally been applied to fMRI data in the visual domain. Compared to the visual domain, current research using RSA for brain decoding in the auditory domain faces two challenges. First, there is a lack of information on how the human brain processes sound. Sound is fundamental to language and communication, and contains both social and biological information. Exploring how the brain processes sound and its characteristics can better understand processes such as sound comprehension and recognition, which is crucial for understanding the functions of the language and auditory systems. Although various multivariate analysis methods have been developed, only a few studies have applied these methods to time-series neuroimaging data to date. Second, existing studies on RSA decoding in the auditory domain have focused on localizing brain regions with distinct response characteristics to stimuli. However, this approach presents numerous challenges, one of which is the ambiguity of interpretations of adjacent time points. If rapid changes in brain activity occur between two time points, fMRI may be unable to distinguish these changes, instead blurring them into a single, overarching pattern of activity. This can lead to underestimation of differences between adjacent time points in similarity analyses, miss information about certain dynamic processes, and limit the interpretation of pattern similarity. Thus, although RSA is a promising technique for analyzing fMRI data, it does not necessarily allow for specific conclusions to be drawn about how information is encoded in the underlying activation patterns, whether the effects are multidimensional or unidimensional, or any other substantive differences. Summary of the Invention

[0007] To solve the above technical problems, the present invention provides a multivariate neural decoding method based on OPM-MEG and RSA. OPM-MEG can provide high-temporal and spatial resolution data on brain activity and provide rich samples of neuronal pattern information. RSA can capture the subtle interactions between different neurons in the brain, analyze the patterns in these data, and reveal the representation of brain activity at different stages. Therefore, the present invention can combine high-resolution neurophysiological data and high-dimensional analysis methods to understand the process of neural representation more finely and comprehensively.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A multivariate neural decoding method based on OPM-MEG and RSA, comprising the following steps:

[0010] Step S1, acquiring resting-state structural MRI data and performing segmentation processing;

[0011] Step S2: Scan the subject wearing the OPM-MEG sensor helmet using a structured light scanner to obtain a scanned image and perform co-registration with the segmented MRI data;

[0012] Step S3: Play auditory stimulation and complete OPM-MEG data acquisition;

[0013] Step S4: pre-processing the collected OPM-MEG data to obtain segmented data;

[0014] Step S5: averaging different trials of the same auditory stimulus at the segmented data level and constructing a neural response pattern vector, and calculating the Pearson correlation of all possible pairwise stimulus pairs at each time point;

[0015] Step S6: obtaining a representation similarity matrix in the time dimension based on the result of step S5, averaging the off-diagonal elements of the representation similarity matrix, and performing step S6 on all subjects to obtain an average result of neural response pattern similarity;

[0016] Step S7, performing a non-parametric Wilcoxon rank sum test at each time point to determine the time points with differences;

[0017] Step S8: Estimate the source activation time series of each subject using MNE-Python;

[0018] Step S9: Project the result of step S8 onto a common brain template and average it among all subjects, and use T-test to determine the activated area.

[0019] In a second aspect, the present invention provides an electronic device comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned multivariate neural decoding method based on OPM-MEG and RSA.

[0020] In a third aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned multivariate neural decoding method based on OPM-MEG and RSA.

[0021] The beneficial effects of the present invention are:

[0022] (1) Direct neurophysiological data with high spatial and temporal resolution: On the one hand, the OPM-MEG used in the present invention can simultaneously measure neuronal activity at multiple locations in the brain with high spatial and temporal resolution, providing rich samples of neuronal pattern information and more finely capturing the information and processes represented by a large number of neuronal groups. Even small angular differences between adjacent dipoles will produce separable field patterns in MEG statistics. At the same time, the propagation of the magnetic field is less affected by the skull and tissue, so MEG has higher positioning accuracy. On the other hand, OPM-MEG directly measures the magnetic activity of neurons and can provide direct observation of brain activity.

[0023] (2) Comprehensive consideration of the activity patterns of a large number of neuronal groups: The RSA adopted in the present invention can simultaneously utilize the spatial and temporal richness of multi-channel data and comprehensively consider multiple brain regions, rather than being limited to the activities of a single brain region. In the representation space, the neural response to each stimulus is represented as a high-dimensional neural response pattern vector. By calculating the distance of the neural response patterns in the representation space, the decoding representation process of the brain can be revealed, thereby more comprehensively understanding the response pattern of the brain under specific tasks or conditions.

[0024] (3) Dynamic evolution of neural decoding: The present invention combines OPM-MEG with RSA, which can not only locate brain regions with different response characteristics to stimuli, but also understand the process-level information contained in these brain regions and reveal the continuous changes in neural response patterns at different stages of the decoding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flow chart of a multivariate neural decoding method based on OPM-MEG and RSA in the present invention;

[0026] Figure 2 A schematic diagram of a similarity matrix representing an embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of the similarity results of neural response patterns according to an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of source localization results of pattern similarity changes according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described below with reference to the accompanying drawings and examples.

[0030] The present invention provides a multivariate neural decoding method based on OPM-MEG and RSA, such as Figure 1 The specific steps are as follows:

[0031] Step S1: Acquire resting-state structural MRI data and perform segmentation processing. The specific implementation steps are as follows:

[0032] Step S1-1, setting MRI scanning parameters through the magnetic resonance imaging console;

[0033] Step S1-2: The subject lies on the scanning bed and fixes the head;

[0034] Step S1-3: monitoring the subject's condition in real time through the built-in monitoring system of the MRI device, and collecting MRI scan image data in a static state;

[0035] Step S1-4: Segment the acquired MRI scan image data to obtain an anatomical label map containing segmentation, surface reconstruction, and morphological measurement information of various brain structures.

[0036] Step S2: Scan the subject wearing the OPM-MEG sensor helmet using a structured light scanner to obtain a scanned image and perform co-registration with the segmented MRI data. The specific implementation steps are as follows:

[0037] Step S2-1: After the subject wears the rigid helmet with the OPM-MEG sensor, a structured light scanner is used to scan the subject's head to obtain a point cloud image of the relative position between the subject's head and the OPM-MEG sensor;

[0038] Step S2-2: Separating the helmet point cloud and the face point cloud obtained in the point cloud image using a region growing method;

[0039] Step S2-3, registering the rigid helmet model and the helmet point cloud based on a random sampling consistency initial registration algorithm;

[0040] Step S2-4, locating the nose tip region in the scanned head point cloud obtained in S2-1 and the nose tip region in the MRI structural image obtained in S1-4 and registering them using a random sampling consistency initial registration algorithm;

[0041] Step S2-5: Based on the iterative closest point algorithm, the registration results obtained in steps S2-3 and S2-4 are finely registered to minimize the sum of squared Euclidean distances between the nose tip area corresponding to the structured light scan and the MRI structural image. The calculation formula is as follows;

[0042] ,

[0043] in, and is a pair of corresponding points between the MRI structural image and the structured light scanning point cloud, with a total of For corresponding points, It is a rotation matrix used to describe the rotation transformation of MRI structural images and structured light scanning point clouds in three-dimensional space. T is a translation matrix used to describe the translation transformation of MRI structural images and structured light scanning point clouds in three-dimensional space.

[0044] Step S3: Play auditory stimulation and complete OPM-MEG data acquisition. The specific implementation steps are as follows:

[0045] Step S3-1: The subject is provided with a helmet equipped with an OPM-MEG sensor and adjusted to a suitable position;

[0046] Step S3-2: Play auditory stimulation and collect data.

[0047] Step S4: pre-process the collected OPM-MEG data to obtain segmented data. The specific implementation steps are as follows:

[0048] Step S4-1: Perform 1-30 Hz offline bandpass filtering on the collected OPM-MEG data, plot the power spectrum density, and check and delete bad tracks based on the power spectrum density;

[0049] Step S4-2, segmenting the data according to the recorded stimulation synchronization trigger signal, extracting the data 100 ms before and 1000 ms after the baseline;

[0050] Step S4-3, performing artifact removal on the data based on independent component analysis;

[0051] Step S4-4: perform visual inspection to determine and remove abnormal segmented data;

[0052] Step S4-5: The segmented OPM-MEG signal obtained after processing in the above steps is as follows:

[0053] ,

[0054] in, is the basic EEG signal, t is the time, is the kth component of the basic EEG signal, Q is the number of components, A Based on magnetic brain signals The instantaneous amplitude of the kth component, Based on magnetic brain signals The instantaneous phase of the kth component, e is a constant, and j is an imaginary unit.

[0055] Step S5: At the segmented data level, average different trials of the same auditory stimulus and construct a neural response pattern vector. Calculate the Pearson correlation of all possible pairwise stimulus pairs at each time point. The specific implementation steps are as follows:

[0056] Step S5-1, averaging different trials of the same sound stimulus to obtain a neural response pattern vector corresponding to the sound stimulus, where the dimension of the neural response pattern vector is the number of OPM-MEG sensors;

[0057] Step S5-2: Calculate the Pearson correlation R between the two sound stimulus pairs, where an R value of 1 represents that the neural response patterns under the two conditions are completely correlated, and an R value of 0 represents that the neural response patterns under the two conditions are completely uncorrelated:

[0058] ,

[0059] in, and are the brain nerve signal intensities measured by all sensors at a certain time t before and after the baseline, that is, the observed values ​​of the neural response pattern vector, and are the average values ​​of brain neural signal intensities measured by all sensors at all times before and after the baseline, respectively. The numerator is the covariance, which measures whether the pattern distribution before and after the baseline tends to be consistent, and the denominator is the product of the standard deviation for normalization;

[0060] Step S5-3: Repeat steps S5-1 and S5-2 for all possible stimulus pairs, and obtain the correlations of all stimulus pairs at the current moment.

[0061] Step S6: Based on the result of step S5, a representation similarity matrix in the time dimension is obtained, and the off-diagonal elements of the representation similarity matrix are averaged within the subject and between subjects to obtain an average result of the similarity of the neural response pattern. The specific implementation steps are as follows:

[0062] Step S6-1, using the correlations of all stimulus pairs at the current moment in step S5-3 to construct a representation similarity matrix at the current moment, wherein the representation similarity matrix is ​​a square matrix symmetric about the diagonal, with a dimension equal to the number of sound stimuli, and each row or column represents the similarity between one stimulus and all stimuli;

[0063] Step S6-2, repeating step S5-3 for each subject at each time point, to obtain a series of representation similarity matrices for each subject along the time axis;

[0064] Step S6-3: Average the off-diagonal elements in the lower left corner of each similarity matrix, calculate the average Pearson correlation R value at each time point, repeat the off-diagonal element averaging operation for all subjects, average these averages across subjects, and obtain a time series of neural response pattern similarity.

[0065] Step S7: Perform non-parametric Wilcoxon rank sum test at each time point to determine the time points with differences. The specific implementation steps are as follows:

[0066] Step S7-1, calculating the average neural response pattern similarity of each subject 100ms before the baseline;

[0067] Step S7-2, combining the average neural response pattern similarity of each subject at a certain moment after the baseline with the average neural response pattern similarity 100 ms before the baseline;

[0068] Step S7-3: sort the merged whole according to the value from small to large;

[0069] Step S7-4, assigning a rank r to each sorted neural response pattern similarity;

[0070] Step S7-5: Use the U statistic as the statistic for the Wilcoxon rank sum test. The calculation formula is:

[0071] ,

[0072] ,

[0073] ,

[0074] in, and are the sample sizes of data at a certain moment before and after the baseline, and Respectively represent the rank of each sample after merging and sorting at a certain moment before and after the baseline;

[0075] Step S7-6: Calculate the rank sum of the pre-baseline and post-baseline data respectively to obtain the Wilcoxon rank sum statistic U, and obtain the corresponding significance level according to the Wilcoxon rank sum test distribution table. value;

[0076] Step S7-7, set the difference level to 0.05, and search the distribution table to find the corresponding rank sum critical value;

[0077] Step S7-8: If the rank sum obtained in S7-6 is greater than the rank sum critical value in S7-7, it is considered that the similarity of the neural response pattern at this moment is different from the baseline; otherwise, it is considered that there is no difference;

[0078] Step S7-9: Repeat step S7-8 at each time point to obtain a series of time points with differences.

[0079] Step S8: Estimate the source activation time series of each subject using MNE-Python. The specific implementation steps are as follows:

[0080] Step S8-1, obtaining the position of the subject's OPM-MEG sensor and the corresponding brain model data from step S2-1;

[0081] Step S8-2, obtaining the subject's source space information from step S1-4;

[0082] Step S8-3, calculating a forward solution by combining the position of the OPM-MEG sensor, the subject's brain model data, and the subject's source space information to obtain the contribution of the current source in the brain to the scalp magnetic field;

[0083] Step S8-4: Based on the dynamic statistical parameter mapping algorithm, estimate the source space activity of the significant time interval obtained in step S7-9, and standardize the result into a Z value.

[0084] Step S9: Project the results of step S8 onto a common brain template and average them across subjects, and use a T-test to determine the activated area. The specific implementation steps are as follows:

[0085] Step S9-1, projecting the source activity Z values ​​of all subjects in step S8-4 onto the common brain template and averaging them;

[0086] Step S9-2: For the Z value obtained in step S9-1, use the T test to calculate the statistical value of each vertex and time point and further obtain the corresponding test significance level. value;

[0087] Step S9-3: To avoid the problem of multiple comparisons, use FDR correction to adjust the significance level of the test. Correction is performed and the significance level of the test is adjusted The value replaces the original data value;

[0088] Step S9-4: The values ​​were projected onto the common brain template and the source activation distribution of the time interval obtained in steps S7-9 was viewed based on the MNE visualization tool.

[0089] Figure 2 The schematic result of representing the similarity matrix is ​​shown. N represents the number of sound stimuli, so the dimension of the matrix is ​​N. 2 Each value in the matrix represents the correlation between the sound stimulus pairs in the corresponding row and column. This is a square matrix that is symmetrical about the diagonal. The closer the value is to 1, the more correlated the response patterns of the two stimuli are. The closer the value is to 0, the less correlated the response patterns of the two stimuli are. Each row or column represents the correlation between the stimulus and all stimuli.

[0090] Figure 3 The figure shows the changes in the similarity of neural response patterns during auditory processing. The curve represents the average similarity value at each time point, reflecting the changes in the similarity of neural response patterns over time. The light shade represents the standard deviation, and the horizontal line above the curve marks the time points that are different from the baseline.

[0091] Figure 4 The sources of the above-mentioned pattern similarity changes were displayed from the lateral, medial and ventral perspectives. It can be seen that the bilateral temporal lobes and parahippocampal gyrus of the brain contribute to the changes in the similarity of neural response patterns.

[0092] In a second aspect, the present invention provides an electronic device comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned multivariate neural decoding method based on OPM-MEG and RSA.

[0093] In a third aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned multivariate neural decoding method based on OPM-MEG and RSA.

[0094] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multivariate neural decoding method based on OPM-MEG and RSA, characterized in that: The following steps are involved: Step S1, acquiring resting-state structural MRI data and performing segmentation processing; Step S2: Scan the subject wearing the OPM-MEG sensor helmet using a structured light scanner to obtain a scanned image and perform co-registration with the segmented MRI data; Step S3: Play auditory stimulation and complete OPM-MEG data acquisition; Step S4: pre-processing the collected OPM-MEG data to obtain segmented data; Step S5: averaging different trials of the same auditory stimulus at the segmented data level and constructing a neural response pattern vector, and calculating the Pearson correlation of all possible pairwise stimulus pairs at each time point; Step S6: obtaining a representation similarity matrix in the time dimension based on the result of step S5, averaging the off-diagonal elements of the representation similarity matrix, and performing step S6 on all subjects to obtain an average result of neural response pattern similarity; Step S7, performing a non-parametric Wilcoxon rank sum test at each time point to determine the time points with differences; Step S8: Estimate the source activation time series of each subject using MNE-Python; Step S9: Project the result of step S8 onto a common brain template and average it among all subjects, and use T-test to determine the activated area.

2. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 1, characterized in that The step S1 comprises: Step S1-1, setting MRI scanning parameters through the magnetic resonance imaging console; Step S1-2: The subject lies on the scanning bed and fixes the head; Step S1-3: monitoring the subject's condition in real time through the built-in monitoring system of the MRI device, and collecting MRI scan image data in a static state; Step S1-4: Segment the acquired MRI scan image data to obtain an anatomical label map containing segmentation, surface reconstruction, and morphological measurement information of various brain structures.

3. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 2, characterized in that: The step S2 comprises: Step S2-1: After the subject wears the rigid helmet with the OPM-MEG sensor, a structured light scanner is used to scan the subject's head to obtain a point cloud image of the relative position between the subject's head and the OPM-MEG sensor; Step S2-2: Separating the helmet point cloud and the face point cloud obtained in the point cloud image using a region growing method; Step S2-3, registering the rigid helmet model and the helmet point cloud based on a random sampling consistency initial registration algorithm; Step S2-4, locating the nose tip region in the point cloud image obtained in step S2-1 and the nose tip region in the MRI scan image obtained in step S1-4, and aligning them using a random sampling consistency initial registration algorithm; Step S2-5: Based on the iterative closest point algorithm, the registration results obtained in steps S2-3 and S2-4 are finely registered to minimize the sum of squared Euclidean distances between the point cloud image and the MRI scan image corresponding to the nose tip area; ， in, and is a pair of corresponding points between the MRI scan image and the point cloud image, with a total of For corresponding points, represents the rotation matrix and T represents the translation matrix.

4. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 3, characterized in that: The S4 includes: Step S4-1: Perform 1-30 Hz offline bandpass filtering on the collected OPM-MEG data, plot the power spectrum density, and check and delete bad tracks based on the power spectrum density; Step S4-2, segmenting the data according to the recorded auditory stimulation synchronization trigger signal, extracting the data 100 ms before and 1000 ms after the baseline; Step S4-3, performing artifact removal based on independent component analysis on the data extracted in step S4-2; Step S4-4: Perform visual inspection, judge and remove abnormal segmented data, and obtain segmented OPM-MEG signal data.

5. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 4, characterized in that: The segmented OPM-MEG signal is: ， in, is the basic EEG signal, t is the time, is the kth component of the basic EEG signal, Q is the number of components, A Based on magnetic brain signals The instantaneous amplitude of the kth component, for The instantaneous phase of the kth component, e is a constant, and j is an imaginary unit.

6. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 5, characterized in that: The S5 includes: Step S5-1, averaging different trials of the same sound stimulus to obtain a neural response pattern vector corresponding to the sound stimulus, where the dimension of the neural response pattern vector is the number of OPM-MEG sensors; Step S5-2: Calculate the Pearson correlation R between the two sound stimulus pairs, where an R value of 1 represents that the neural response patterns under the two conditions are completely correlated, and an R value of 0 represents that the neural response patterns under the two conditions are completely uncorrelated: ， in, and are the brain nerve signal intensities measured by all sensors at time t before and after the baseline, i.e., the observed values ​​of the neural response pattern vector, and are the average values ​​of brain neural signal intensities measured by all OPM-MEG sensors at all times before and after baseline, respectively; Step S5-3: Repeat steps S5-1 and S5-2 for all possible pairs of sound stimuli to obtain the correlations of all stimulus pairs at the current moment.

7. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 6, characterized in that: The S6 includes: Step S6-1, using the correlations of all stimulus pairs at the current moment in step S5-3 to construct a representation similarity matrix at the current moment, wherein the representation similarity matrix is ​​a square matrix symmetric about the diagonal, with a dimension equal to the number of sound stimuli, and each row or column represents the similarity between one stimulus and all stimuli; Step S6-2, repeating step S5-3 for each subject over time, to obtain a series of representation similarity matrices for each subject along the time axis; Step S6-3: Average the off-diagonal elements in the lower left corner of each similarity matrix, calculate the average Pearson correlation R value at each time point, repeat the off-diagonal element averaging operation for all subjects, average the obtained average values ​​across subjects, and obtain a time series of neural response pattern similarity.

8. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 7, characterized in that: The S7 includes: Step S7-1, calculating the average neural response pattern similarity of each subject 100ms before the baseline; Step S7-2, combining the average neural response pattern similarity of each subject at a certain moment after the baseline with the average neural response pattern similarity 100 ms before the baseline into a whole; Step S7-3: sort the merged whole according to the value from small to large; Step S7-4, assigning a rank r to each sorted neural response pattern similarity; Step S7-5: Use the U statistic as the statistic for the Wilcoxon rank sum test. The calculation formula is: ， ， ， in, and are the sample sizes of data at a certain moment before and after the baseline, and Respectively represent the rank of each sample after merging and sorting at a certain moment before and after the baseline; Step S7-6: Calculate the rank sum of the pre-baseline and post-baseline data respectively to obtain the Wilcoxon rank sum statistic U, and obtain the corresponding significance level according to the Wilcoxon rank sum test distribution table. value; Step S7-7, set the difference level to 0.05, and search the distribution table to find the corresponding rank sum critical value; Step S7-8: If the rank sum obtained in step S7-6 is greater than the rank sum critical value in step S7-7, it is considered that the similarity of the neural response pattern at this moment is different from the baseline; otherwise, it is considered that there is no difference; Step S7-9: Repeat step S7-8 at each time point to obtain a series of time points with significant differences.

9. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 8, characterized in that: The S8 includes: Step S8-1, obtaining the position of the subject's OPM-MEG sensor and the corresponding brain model data from step S2-1; Step S8-2, obtaining the subject's source space information from step S1-4; Step S8-3, calculating a forward solution by combining the position of the OPM-MEG sensor, the subject's brain model data, and the subject's source space information to obtain the contribution of the current source in the brain to the scalp magnetic field; Step S8-4: Based on the dynamic statistical parameter mapping algorithm, estimate the source space activity of the significant difference time points obtained in step S7-9, and standardize the results into Z values.

10. The multivariate neural decoding method based on OPM-MEG and RSA according to claim 9, characterized in that: The S9 includes: Step S9-1, project the Z values ​​corresponding to the source activities of all subjects in S8-4 onto the common brain template and average them; Step S9-2: Use T-test to calculate the statistical value of each vertex and time point and obtain the corresponding test significance level value; Step S9-3, test significance level The significance level of the corrected test was obtained after the value was corrected. value, replacing the significance level in the original data value; Step S9-4: Set the corrected test significance level of step S9-3 to The values ​​were projected onto a common brain template and the spatial distribution of source activations at the time points obtained in steps S7-9 was viewed using MNE-based visualization tools.

11. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs; When one or more programs are executed by the one or more processors, the one or more processors implement the multivariate neural decoding method based on OPM-MEG and RSA according to any one of claims 1 to 10.

12. A computer-readable storage medium, characterized in that Executable instructions are stored thereon, which, when executed by a processor, enable the processor to implement the multivariate neural decoding method based on OPM-MEG and RSA as described in any one of claims 1 to 10.

Images