A method for analyzing connectivity of auditory representations based on spatiotemporal information

Abstract

The application discloses a kind of based on spatiotemporal information's auditory representation connectivity analysis method, comprising: obtaining MRI data and carrying out segmentation processing;Play auditory stimulation and complete fMRI data acquisition;Based on laser scanner, the subject who has worn OPM-MEG sensor helmet is scanned, obtains scanning image and is jointly registered with MRI data;Brain region activation level analysis is carried out to fMRI data, and the activated brain region ROI under task state is obtained;According to ROI and registration result, the OPM-MEG sensor corresponding to ROI is screened out;Play auditory stimulation and complete OPM-MEG data acquisition;OPM-MEG data is preprocessed to obtain basic OPM-MEG signal;Granger causality analysis is carried out to basic OPM-MEG signal;The result obtained is compared between subjects and statistical test is carried out to detect whether there is difference between the neural representation processes of subjects.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of biomedical engineering, and particularly relates to a hearing representation connectivity analysis method based on space-time information. BACKGROUND

[0002] Sound has a profound impact on many organisms, including humans, in addition to containing basic language information, it also carries information in multiple aspects such as social interaction, culture, learning, emotional expression and survival, and is an indispensable part of social interaction, cultural heritage and individual survival. Therefore, accurately perceiving this information is crucial.

[0003] Hearing neural representation disorder refers to difficulties in processing auditory information, which affects an individual's ability to perceive, understand and process sound and language. This may involve various levels of the auditory system, including the ear, auditory nerve, auditory cortex, etc.

[0004] Hearing neural representation refers to the way the brain encodes and processes sound. When the auditory nervous system receives sound signals, it converts these signals into neuronal activity and forms specific representations in the brain for the perception and understanding of sound. Hearing neural representation disorder is closely related to hearing neural representation. Hearing neural representation determines to some extent how the brain understands and processes sound, and hearing neural representation disorder can affect the formation and function of these neural representations. Hearing neural representation is the basis for understanding and processing sound, and hearing neural representation disorder can affect these foundations, therefore, understanding hearing neural representation is of great significance for understanding the mechanism of hearing neural representation disorder.

[0005] Functional magnetic resonance imaging (fMRI) is a neuroimaging technique used to study brain activity. It is based on magnetic resonance imaging (MRI) technology and measures blood flow and metabolic rate in different regions of the brain, revealing the activity of the brain when performing specific tasks or in certain states. The principle behind fMRI is based on the blood oxygen level-dependent (BOLD) signal. When a certain brain area is active, the increased neuronal activity in that region leads to the dilation of surrounding blood vessels and an increase in local blood flow to meet the energy demands of the active area. Changes in blood flow result in changes in the proportion of oxygenated and deoxygenated hemoglobin in the blood, which in turn alters the distribution of the local magnetic field, producing the BOLD signal. Although fMRI measures changes in blood flow, this change is closely related to neuronal activity. Typically, neuronal activity and local blood flow increase are synchronized, so the BOLD signal can be considered an indicator of brain activity. When performing an fMRI scan, the subject is usually placed in a large magnetic field, which is typically generated by a superconducting magnet. In the presence of a large magnetic field, the protons in the human body naturally precess along the direction of the magnetic field, a phenomenon known as precession resonance. When given the appropriate radio frequency pulse, protons absorb and release energy, and this energy release is called resonance signal. By measuring these signals, high-resolution images of human tissues can be obtained. Blood needs some time to be transported to the active brain area, resulting in a delay in the change of blood flow. This delay results in a certain time difference between the signal response measured by fMRI and the actual neuronal activity; in addition, it takes some time to acquire fMRI data at each time point, including exciting magnetic resonance signals, data reception and processing, etc. In order to obtain a series of data at different time points, a considerable amount of time is required, and fMRI data usually requires complex signal processing and statistical analysis to determine the activity level of different brain regions, which requires additional time. Due to these reasons, although fMRI has high spatial resolution, its temporal resolution is relatively low.

[0006] Magnetoencephalography (MEG) is a non-invasive neuroimaging technique used to measure the magnetic fields generated by brain activity. When neurons are active, ions move across the neuron membrane, causing a change in the electric potential difference between the inside and outside of the cell. This change in electric potential difference causes a current to flow, forming a weak current loop. According to Ohm's law, electric current generates a magnetic field. Therefore, the weak current loop generated by neuronal electrical activity also generates a weak magnetic field, which is the aforementioned MEG signal. Although MEG signals are very weak, they can be detected by sensitive devices such as optically pumped magnetometers (OPMs).

[0007] OPM is a highly sensitive magnetic sensor that uses optical methods to measure magnetic fields. At the atomic scale, particles exhibit quantum properties, including the spins of atomic nuclei and electrons, thus generating a tiny magnetic moment that makes atoms tiny magnetic dipoles. OPM uses laser to excite atomic nuclei or electrons in atoms, causing the spin direction to change. In this process, the atom absorbs energy and transitions from the ground state to a high-energy level. When an external magnetic field is superimposed, it will affect the energy level structure of the atom. This external magnetic field will cause the splitting of atomic energy levels, one of which is called the Larmor splitting. The frequency difference between the split energy levels is called the Larmor frequency. When the external magnetic field changes, the Larmor frequency of the atom changes, and the laser absorption spectrum also changes. OPM uses this principle to manipulate atomic spins with laser and measures the laser absorption spectrum to detect changes in the magnetic field, so it can accurately measure the changes in the weak magnetic field.

[0008] Currently, the main approaches to understanding auditory neural representation include auditory tests, neuroimaging examinations, and behavioral tests. Although these methods play an important role in auditory neural representation, there are still some problems. First, the current evaluation methods are highly subjective, which may lead to significant differences in the understanding of the representation process by different institutions and researchers. Second, MRI and computed tomography (CT) can only be used to examine the structure of the central nervous system, and more information about the structure of the brain is detected, but the functional information of the brain is lacking. In addition, the time resolution of fMRI is usually on the order of seconds, which cannot capture the rapid dynamic changes in the auditory representation process. If there is a rapid change in brain activity between two time points, fMRI may not be able to distinguish this change and blur it as a whole activity pattern, but in fact, auditory processing is a rapid and dynamic process, so fMRI may miss some important information on the time scale.

[0009] MEG can provide high time resolution brain activity images and accurately locate the source and time of brain activity. Combined with the high sensitivity of OPM, it can detect subtle changes in brain activity more finely based on MEG, thus providing more comprehensive and accurate brain activity images. Therefore, the combination of OPM-MEG and fMRI for auditory neural representation can fully utilize their complementarity. In addition, OPM-MEG and fMRI provide different types of neurophysiological data, such as BOLD signals and neuronal electrical activity. Combining these two types of information can provide a more comprehensive understanding of brain function and activity. SUMMARY

[0010] To address the shortcomings of current methods for auditory neural representation, such as subjective assessment, limited neurophysiological data, and low temporal resolution, this invention proposes a spatiotemporal information-based auditory representation connectivity analysis method, which enables a more objective and scientific understanding of the auditory neural representation process.

[0011] To achieve the above objectives, the present invention provides the following solution:

[0012] A method for auditory representation connectivity analysis based on spatiotemporal information includes the following steps:

[0013] S1. Acquire resting-state structural MRI data and perform segmentation processing;

[0014] S2. Play auditory stimuli and complete fMRI data acquisition;

[0015] S3. The subjects wearing OPM-MEG sensor helmets are scanned using a laser scanner to obtain scan images and perform joint registration with the segmented resting-state structural MRI data;

[0016] S4. Perform brain region activation level analysis on the fMRI data under task mode to obtain the brain regions activated under task mode. The brain regions are the regions of interest (ROIs) for subsequent analysis.

[0017] S5. Based on the region of interest (ROI) and registration results, select the OPM-MEG sensor corresponding to the ROI.

[0018] S6. Play auditory stimuli and complete OPM-MEG data acquisition;

[0019] S7. Preprocess the OPM-MEG data to obtain the basic OPM-MEG signal and perform time-frequency analysis. Calculate the average Z_score waveform and take the maximum value of the waveform as the response intensity.

[0020] S8. Perform Granger causality analysis on the basic OPM-MEG signal;

[0021] S9. The results obtained by the subjects in S7 and S8 were compared and statistically tested among the subjects to detect whether there were differences in the neural representation process among the subjects.

[0022] Preferably, in step S1, the method for acquiring and segmenting resting-state structural MRI data includes:

[0023] S1-1. Set MRI scan parameters using the MRI console and ensure the subject remains in the correct position during the scan to obtain MRI scan data;

[0024] S1-2, segmenting the MRI scan data using the software tool FreeSurfer to segment different tissue structures in the MRI scan data into respective regions to obtain segmentation results containing each tissue structure in the MRI scan data.

[0025] Preferably, in the S2, the method of playing the auditory stimulus and completing the fMRI data acquisition comprises:

[0026] S2-1, setting fMRI scan parameters through a magnetic resonance imaging console;

[0027] S2-2, editing the auditory stimulus based on the fMRI scan parameters using Praat;

[0028] S2-3, controlling the edited auditory stimulus using MATLAB;

[0029] S2-4, presenting the auditory stimulus using Psychphysics Toolbox 3.0 to complete the fMRI data acquisition.

[0030] Preferably, in the S3, the method of scanning the subject wearing the OPM-MEG sensor helmet based on the laser scanner, obtaining the scan image, and jointly registering with the segmented resting-state structural MRI data comprises:

[0031] S3-1, scanning the subject wearing the sensor helmet based on the laser scanner to obtain a scan image;

[0032] S3-2, based on the scan image, using the region growing method to separate the helmet point cloud from the face point cloud;

[0033] S3-3, performing coarse registration of the helmet model and the separated helmet point cloud based on the random sample consensus initial registration algorithm SAC-IA;

[0034] S3-4, further fine registration of the point cloud after S3-3 coarse registration based on the iterative closest point algorithm ICP;

[0035] S3-5, positioning the nose tip point of the face point cloud and the resting-state structural MRI data and performing coarse registration;

[0036] S3-6, further matching the point cloud after S3-5 coarse registration based on the iterative closest point algorithm ICP, completing the registration of the laser scanning result and the resting-state structural MRI data, and obtaining the position and direction information of the OPM-MEG sensor relative to the human brain.

[0037] Preferably, in the S4, the method of analyzing the brain region activation level of the fMRI data under the task state to obtain the activated brain region under the task state, which is the region of interest ROI for subsequent analysis comprises:

[0038] S4-1, using a general linear model GLM to analyze the brain activation level of the task-state fMRI data, the calculation formula is as follows:

[0039] Y = βX + ε, wherein Y represents the BOLD signal of each voxel, X represents some unknown numbers, β represents a linear combination, and ε represents a residual error;

[0040] S4-2, using the least square method to minimize the square sum of the residual error ε to fit the parameter β value;

[0041] S4-3, statistically inferring the β value and setting a predetermined threshold to obtain the brain function activation map corresponding to the sound stimulation, i.e. the region of interest ROI for subsequent analysis.

[0042] Preferably, in the S6, the method of playing the auditory stimulus and completing the OPM-MEG data acquisition comprises:

[0043] S6-1, wearing a helmet with OPM-MEG sensors for the subject and adjusting to a suitable position;

[0044] S6-2, playing the auditory stimulus and collecting data.

[0045] Preferably, in the S7, the method of pre-processing the OPM-MEG data to obtain the basic OPM-MEG signal and performing time-frequency analysis, calculating the average Z_score waveform, and taking the maximum value of the waveform as the response intensity comprises:

[0046] S7-1, performing 1-30Hz offline band-pass filtering and drawing power spectral density to check and delete bad channels;

[0047] S7-2, according to the recorded stimulus synchronous trigger signal, data segmentation is performed, and data 100ms before stimulation and 1000ms after stimulation are extracted;

[0048] S7-3, visual inspection is performed to judge and remove abnormal segmented data obtained in S7-2 to obtain the basic OPM-MEG signal;

[0049] S7-4, using continuous wavelet transform CWT based on Morlet wavelet to generate the time-frequency distribution of the basic OPM-MEG signal to obtain the corresponding time-frequency graph;

[0050] S7-5, normalizing the time-frequency graph, and calculating Z_score at each time point to obtain the Z_score waveform, and defining the maximum value of the Z_score waveform as the response intensity of the subject to the sound stimulus.

[0051] Preferably, in S8, the method of performing Granger causality analysis on the basic OPM-MEG signal comprises:

[0052] S8-1, constructing a first autoregressive model with Y(t) as the dependent variable and past values of Y(t) as the independent variable;

[0053] S8-2, constructing a second autoregressive model with Y(t) as the dependent variable and past values of Y(t) and X(t) as the independent variable;

[0054] S8-3, fitting the residuals of the first autoregressive model and the second autoregressive model respectively, and calculating the variances of the residual terms;

[0055] S8-4, if X(t) is the Granger cause of Y(t), then the variance of the residual term of the second autoregressive model should be smaller than the variance of the residual term of the first autoregressive model, and the classical time-domain Granger causality index is calculated by the variance of the residual term of the first autoregressive model and the variance of the residual term of the second autoregressive model;

[0056] S8-5, extracting characteristic statistics in the Granger causality index to evaluate the causal relationship between time series;

[0057] S8-6, according to the extracted characteristic statistics and the time-domain Granger causality index, the information flow direction in the auditory nerve representation is drawn.

[0058] Preferably, in S9, the method of comparing and statistically testing the results obtained by the subjects in S7 and S8 to detect whether there is a difference in the neural representation process between the subjects comprises:

[0059] S9-1, using U statistics as the statistic of Wilcoxon rank sum test;

[0060] S9-2, defining the null hypothesis as the difference between the response intensity statistics of the subject and others;

[0061] S9-3, calculating the U statistics of the response intensity at each time point, and sorting the values of the U statistics according to the size;

[0062] S9-4, assigning a rank R to each value;

[0063] S9-5, adding the rank corresponding to each value to obtain the total rank sum U1, U2, and finally obtaining the U statistics based on the total rank sum U1, U2;

[0064] S9-6, selecting P=0.05 as the significance level, and finding the corresponding rank sum critical value according to the Wilcoxon rank sum test distribution table;

[0065] S9-7. If the rank sum obtained in S9-5 is greater than the critical value of the rank sum in S9-6, then reject the null hypothesis, that is, consider that the response strength is different from others; otherwise, accept the null hypothesis, that is, consider that the response strength is not different from others.

[0066] S9-8: Based on the comparisons among subjects in S7 and the statistical test results in S9, we obtain the specific details of the subjects' auditory neural representation; based on the comparisons among subjects in S8 and the statistical test results in S9, we obtain the information flow and corresponding brain regions in the process of the subjects' auditory neural representation.

[0067] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0068] (1) More multidimensional and direct neurophysiological data: The OPM-MEG and fMRI used in this invention are two independent and different brain activity measurement techniques. On the one hand, compared with the fMRI method that is currently widely used for auditory nerve representation, since its essential measurement principle is to infer neural activity by measuring BOLD signals, it can only provide indirect neural activity data. OPM-MEG measures the magnetic field changes generated by the electrical activity of neurons, providing direct physiological data of neuronal activity. On the other hand, as mentioned above, OPM-MEG and fMRI are two different modal measurement techniques, so cross-modal verification can be performed to enhance the credibility of the results.

[0069] (2) A more objective analysis process: Currently, the methods widely used for auditory nerve representation, such as hearing tests, behavioral and cognitive tests, questionnaires and interviews, are easily affected by various subjective factors such as the psychology and physiology of the subjects, and differences in research institutions and personnel. The ability to make accurate assessments is limited, resulting in a high rate of misjudgment and incorrect judgment of auditory nerve representation and an inability to understand the specific neural mechanisms of individuals. The method adopted in this invention reduces the influence of subjective factors, is more objective, and can understand the specific auditory representation process of the subjects.

[0070] (3) Consideration of individual differences: Due to the existence of individual differences, the brain activation of each subject is not exactly the same. The fMRI method used in this invention provides important prior information for OPM-MEG, which not only improves the specificity of the measurement but also improves the efficiency of data analysis. This is of great significance for individualized research.

[0071] (4) Multi-dimensional comprehensive utilization of time-domain, spatial-domain, and frequency-domain information: On the one hand, the OPM-MEG in this invention provides high temporal resolution, which can accurately capture the rapid dynamic changes of neural activity, while fMRI provides high spatial resolution, which can locate the precise spatial position of neural activity. The combined analysis can provide a more comprehensive perspective in time and space and reveal auditory neural representations more precisely. On the other hand, the OPM-MEG used in this invention has good sensitivity to spectral information, so it can analyze neural activity in different frequency bands. By combining it with fMRI, it can more comprehensively understand the specific situation of neural representations in different frequency domains. Attached Figure Description

[0072] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a flowchart of an auditory representation connectivity analysis method based on spatiotemporal information according to an embodiment of the present invention;

[0074] Figure 2 This is a schematic diagram of the joint registration results in an embodiment of the present invention;

[0075] Figure 3 This is a schematic diagram showing the response intensity and Wilcoxon rank-sum test results of an embodiment of the present invention;

[0076] Figure 4 This is a schematic diagram of the Granger causality analysis results in an embodiment of the present invention. Detailed Implementation

[0077] The technical solutions of the embodiments 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0079] Example 1

[0080] This invention provides a method for auditory representation connectivity analysis based on spatiotemporal information, such as... Figure 1 As shown, the specific steps are as follows:

[0081] S1. Acquire resting-state structural MRI data and segment it for registration in S3. The specific steps are as follows:

[0082] S1-1. The operator sets the MRI scan parameters through the MRI console and ensures that the subject remains in the correct position during the scan;

[0083] S1-2. Use the existing software tool FreeSurfer (https: / / surfer.nmr.mgh.harvard.edu / ) to segment the obtained MRI scan data, dividing different tissue structures in the MRI image into their respective regions. For example, brain tissue is segmented into the cerebral cortex, gray matter, white matter, cerebrospinal fluid, etc. Finally, FreeSurfer generates a detailed anatomical label map, which contains the segmentation results of each tissue structure in the MRI image.

[0084] S2. Play auditory stimuli and complete fMRI data acquisition. The specific steps are as follows:

[0085] S2-1. The operator sets the fMRI scan parameters through the magnetic resonance imaging console;

[0086] S2-2. Use Praat (https: / / www.fon.hum.uva.nl / praat / ) to edit the auditory stimuli needed according to the research requirements;

[0087] S2-3. Use MATLAB (https: / / www.mathworks.com / products / matlab.html) to control the edited auditory stimuli;

[0088] S2-4. Use Psychphysics Toolbox 3.0 (http: / / psychtoolbox.org / ) to present stimuli.

[0089] S3. Scan the subject wearing the OPM-MEG sensor helmet using a laser scanner to obtain the scanned images and perform joint registration with the MRI scan data obtained in S1. The specific steps are as follows:

[0090] S3-1. Using a laser scanner, scan the subject wearing the sensor helmet to obtain scan images;

[0091] S3-2. Use the existing region growing method to separate the helmet point cloud from the face point cloud;

[0092] S3-3. Based on the existing Random Sample Consensus Initial Registration (SAC-IA) algorithm, coarse registration is performed on the helmet model and the helmet point cloud obtained from S3-2.

[0093] S3-4. Based on the existing Iterative Closest Point (ICP) algorithm, further refine the point cloud after coarse registration in S3-3. The core of the ICP algorithm is to minimize an objective function, which is actually the sum of squares of the Euclidean distances between all corresponding points. The formula for calculating the objective function is as follows:

[0094]

[0095] in, and There are N corresponding points between the target point cloud and the source point cloud. p For a given point, R is a rotation matrix used to describe the rotation transformation of the point cloud in three-dimensional space, and T is a translation matrix used to describe the translation transformation of the point cloud in three-dimensional space.

[0096] S3-5. Locate the nasal tip point of the face point cloud obtained in S3-2 and the MRI structural image obtained in S1 and perform coarse registration.

[0097] S3-6. Based on the ICP algorithm, the point cloud after coarse registration in S3-5 is further matched to complete the registration of laser scanning results with MRI structural images, and obtain the position and orientation information of the OPM-MEG sensor relative to the human brain.

[0098] S4. Perform brain region activation level analysis on the task-oriented fMRI data to obtain the activated brain regions under task conditions. These brain regions are the regions of interest (ROIs) for subsequent analysis. The specific principle is as follows:

[0099] S4-1. A general linear model (GLM) is used to analyze the activation levels of brain regions in task-oriented fMRI data. GLM assumes that the BOLD signal (represented by Y) on each voxel is a linear combination (represented by β) of some unknowns (represented by X). The matrix composed of X is also called the design matrix, and the calculation formula is as follows:

[0100] Y = βX + ε

[0101] Sound stimuli are not directly used as independent variables, but are convolved with the hemodynamic response function (HRF) and used as regression factors of interest; regression factors of no interest, such as head movement parameters, are directly used as covariates; the corresponding β value represents the correlation between the influencing factors and brain region activation, and the larger the β value, the stronger the correlation.

[0102] S4-2, ε represents the residuals. The least squares method is used to minimize the sum of squares of the residuals to fit the parameter β value.

[0103] S4-3. Perform statistical inference on the β value, usually by performing a t-test or F-test and setting a certain threshold, to obtain the brain function activation map corresponding to the sound stimulus in S2-4, thus obtaining the ROI.

[0104] S5. Based on the registration results of the ROI obtained in S4 and S3, the OPM-MEG sensor corresponding to the ROI is selected. The specific principle is as follows:

[0105] Since the OPM-MEG sensor helmet is a custom-made helmet, the position and orientation of the sensor relative to the helmet are known. After the two-step registration in S3-5 and S3-6, the position and orientation of the sensor relative to the subject's MRI can be obtained. Combined with the ROI obtained in S4-3, the OPM-MEG sensor corresponding to the ROI region and its position and orientation can be obtained.

[0106] S6. Play auditory stimuli and complete OPM-MEG data acquisition. The specific steps are as follows:

[0107] S6-1. Put the helmet containing the OPM-MEG sensor on the subject and adjust it to a suitable position;

[0108] S6-2, Play auditory stimuli and collect data.

[0109] S7. Preprocess the initial OPM-MEG data obtained in S6 to obtain the basic OPM-MEG signal and perform time-frequency analysis. Calculate the average Z_score waveform and take the maximum value of the waveform as the response intensity. The specific steps are as follows:

[0110] S7-1, Perform 1-30Hz offline bandpass filtering and plot the power spectral density to check and remove bad channels;

[0111] S7-2. Based on the recorded stimulus synchronization trigger signal, perform data segmentation and extract data from 100ms before the stimulus and 1000ms after the stimulus for subsequent analysis.

[0112] S7-3. Conduct a visual inspection to identify and remove abnormal segmented data obtained in S7-2;

[0113] S7-4. The calculation formula for the basic OPM-MEG signal obtained after the above steps is as follows:

[0114]

[0115] Where x(t) is the basic magnetoencephalogram (MEG) signal, t is time, and x... k (t) is the k-th component, N is the number of components, and Am k (t) is the instantaneous amplitude of the k-th component of x(t). Let be the instantaneous phase of the k-th component of x(t), e be a constant, and j be the imaginary unit;

[0116] S7-5. Use the Continuous Wavelet Transform (CWT) based on Morlet wavelets to generate the time-frequency distribution of the basic OPM-MEG signal x(t) in S7-4, and obtain the corresponding time-frequency plot. The calculation formula for the Morlet wavelet is:

[0117]

[0118] Where ω is the center frequency of the Morlet wavelet, and σ is the control Gaussian kernel. The parameters are e, a constant, and j, the imaginary unit.

[0119] The formula for calculating the CWT of signal x(t) is:

[0120]

[0121] Where α is a scaling factor controlling the wavelet's scale in the time domain, and τ is a translation factor controlling the wavelet's time-domain shift. yes Wavelet basis obtained by scaling and translation in the time domain;

[0122] S7-6. Normalize the time-frequency graph obtained in S7-5, and calculate the Z_score for each time point to obtain the Z_score waveform. The maximum value of the Z_score waveform is defined as the intensity of the subject's response to the sound stimulus. The formula for calculating Z_score is:

[0123]

[0124] Where μ is the mean of Z_score and σ is the standard deviation of Z_score;

[0125] S8. Perform Granger causality analysis on the basic OPM-MEG signal obtained in S7. The specific steps are as follows:

[0126] S8-1. Construct the first autoregressive model with Y(t) as the dependent variable and past values ​​of Y(t) as the independent variable:

[0127]

[0128] Where Y(tp) represents the value at the p-th time point before time t, and p is the order of the autoregressive model. This parameter needs to be estimated before fitting the autoregressive model, and the Akaike Information Criterion (AIC) is used as the estimation criterion. This represents the residual when predicting Y(t) using past values ​​of Y(t);

[0129] S8-2. Construct a second autoregressive model with Y(t) as the dependent variable and the past values ​​of Y(t) and X(t) as independent variables:

[0130]

[0131] In the two models above, a i (i = 1, 2, ..., p), b i (i = 1, 2, ..., p), c i (i = 1, 2, ..., p) are model parameters, which can be obtained using the least squares method;

[0132] S8-3. In the two models mentioned above... The residuals are the model residuals. After fitting the S8-1 and S8-2 models, the variance of the residual terms is calculated:

[0133]

[0134] S8-4. If X(t) is a Granger cause of Y(t), then It should be smaller than pass and The classic time-domain Granger causality index is calculated using the following formula:

[0135]

[0136] The value range of this indicator is [0, +∞). If GC... X→Y If the value is equal to 0, then past information about X(t) cannot improve the prediction accuracy of Y(t); if GC... X→Y If the value is greater than 0, then X(t) is a Granger cause of Y(t);

[0137] S8-5. Extract statistical measures such as causal strength and lag from the Granger causality analysis obtained in S8-4 to evaluate the causal relationship between time series.

[0138] S8-6. Based on the statistics extracted in S8-5 and the calculation indicators in S8-4, plot the information flow in the auditory nerve representation.

[0139] S9. Compare and statistically test the results obtained by the subjects in S7 and S8 to detect whether there are significant differences in the neural representation processes among the subjects. Taking response intensity as an example, the specific steps are as follows:

[0140] S9-1. The U statistic is used as the statistic for the Wilcoxon rank-sum test. The calculation formula is as follows:

[0141] U = min(U1, U2),

[0142]

[0143] Where n1 and n2 are the sample sizes of the two subjects, R i and R j These represent the rank of each sample in the combined and sorted data of the two subjects;

[0144] S9-2, Define the null hypothesis as follows: The subject's response intensity statistic is not significantly different from that of others;

[0145] S9-3. Calculate the U statistic of the response intensity at each time point and sort its values ​​from smallest to largest.

[0146] S9-4. Assign a rank R to each value, starting from 1 and gradually increasing. For the same value, the average rank can be used.

[0147] S9-5. Sum the corresponding ranks to obtain the total rank sums U1 and U2, and select the smaller of the two as the final U statistic;

[0148] S9-6. Select P=0.05 as the significance level, and find the corresponding critical value of rank sum according to the Wilcoxon rank sum test distribution table;

[0149] S9-7. If the rank sum obtained in S9-5 is greater than the critical value of the rank sum in S9-6, then reject the null hypothesis, that is, consider the response strength to be significantly different from others; otherwise, accept the null hypothesis, that is, consider the response strength to be not significantly different from others.

[0150] S9-8, the comparisons among subjects in S7 and the statistical test results in S9 can reveal the specific details of the subjects' auditory neural representation; the comparisons among subjects in S8 and the statistical test results in S9 can reveal the information flow and corresponding brain regions during the subjects' auditory neural representation process.

[0151] Figure 2 The final registration results are shown, with arrows indicating the position and orientation of the sensors, which have been converted to the subject's MRI coordinate system.

[0152] Figure 3This study demonstrates the intensity of subjects' responses to different types of stimuli and the results of the Wilcoxon rank-sum test. Referring to several existing studies, this invention sets up six categories of sound stimuli: human language (e.g., "good," "hard"), human emotional sounds (e.g., sadness, anger), animal sounds (e.g., dog barks, pig grunts), natural environmental sounds (e.g., wind, rain), everyday scenes (e.g., sirens, clinking glasses), and musical instruments (e.g., piano, violin). Each category contains 15 specific sound stimuli. Using Praat software, the stimuli were set to have the same duration and the volume was normalized to the same amplitude level. All stimuli were presented in a random order with the volume adjusted to a comfortable level, approximately 65 dB sound pressure level. Experimental results show that subjects exhibit significant differences in their responses to different types of stimuli.

[0153] Figure 4 The Granger causality analysis diagram of the subjects' auditory representation is shown, where black dots represent electrode locations, letters represent electrode names, and arrows represent the flow of information after Granger causality analysis. The results show the flow of information between different brain regions during the auditory representation process.

[0154] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0155] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for analyzing auditory representation connectivity based on spatiotemporal information, characterized in that, Includes the following steps: S1. Acquire resting-state structural MRI data and perform segmentation processing; S2. Play auditory stimuli and complete fMRI data acquisition; S3. The subjects wearing OPM-MEG sensor helmets are scanned using a laser scanner to obtain scan images and perform joint registration with the segmented resting-state structural MRI data; S4. Perform brain region activation level analysis on the fMRI data under task mode to obtain the brain regions activated under task mode. The brain regions are the regions of interest (ROIs) for subsequent analysis. S5. Based on the region of interest (ROI) and registration results, select the OPM-MEG sensor corresponding to the ROI. S6. Play auditory stimuli and complete OPM-MEG data acquisition; S7. Preprocess the OPM-MEG data to obtain the basic OPM-MEG signal and perform time-frequency analysis. Calculate the average Z_score waveform and take the maximum value of the waveform as the response intensity. S8. Perform Granger causality analysis on the basic OPM-MEG signal; S9. The results obtained by the subjects in S7 and S8 were compared and statistically tested among the subjects to detect whether there were differences in the neural representation process among the subjects. In step S3, the method of scanning a subject wearing an OPM-MEG sensor helmet using a laser scanner to obtain scanned images and jointly registering them with segmented resting-state structural MRI data includes: S3-1. Using a laser scanner, scan the subject wearing the sensor helmet to obtain scan images; S3-2. Based on the scanned image, the helmet point cloud and the face point cloud are separated using the region growing method; S3-3. The helmet model and the separated helmet point cloud are coarsely registered based on the random sample consistency initial registration algorithm SAC-IA. S3-4. Further fine registration of the point cloud after coarse registration in S3-3 is performed based on the Iterative Closest Point (ICP) algorithm. S3-5. Locate the nasal tip point of the face point cloud and the resting-state structural MRI data and perform coarse registration; S3-6. Based on the Iterative Closest Point (ICP) algorithm, the point cloud after coarse registration in S3-5 is further matched to complete the registration of the laser scanning results with the resting-state structural MRI data, and obtain the position and orientation information of the OPM-MEG sensor relative to the human brain. In S8, the method for performing Granger causality analysis on the basic OPM-MEG signal includes: S8-1. Construct the first autoregressive model with Y(t) as the dependent variable and Y(t) as the past value of the independent variable; S8-2. Construct a second autoregressive model with Y(t) as the dependent variable and the past values ​​of Y(t) and X(t) as independent variables; S8-3. Fit the residuals of the first and second autoregressive models respectively, and calculate the variance of the residual terms. S8-4. If X(t) is a Granger cause of Y(t), then the variance of the residual term of the second autoregressive model should be less than the variance of the residual term of the first autoregressive model. The classic time-domain Granger causality index is calculated by using the variance of the residual term of the first autoregressive model and the variance of the residual term of the second autoregressive model. S8-5. Extract characteristic statistics from Granger causality indicators to assess causal relationships between time series. S8-6. Based on the extracted feature statistics and the time-domain Granger causality index, plot the information flow in the auditory nerve representation.

2. The auditory representation connectivity analysis method based on spatiotemporal information according to claim 1, characterized in that, In step S1, the method for acquiring and segmenting resting-state structural MRI data includes: S1-1. Set MRI scan parameters using the MRI console and ensure the subject remains in the correct position during the scan to obtain MRI scan data; S1-2. Use the software tool FreeSurfer to segment the MRI scan data, dividing the different tissue structures in the MRI scan data into their respective regions, and obtaining the segmentation results containing each tissue structure in the MRI scan data.

3. The auditory representation connectivity analysis method based on spatiotemporal information according to claim 1, characterized in that, In step S2, the method for playing auditory stimuli and completing fMRI data acquisition includes: S2-1. Set fMRI scan parameters via the magnetic resonance imaging console; S2-2. Based on fMRI scan parameters, auditory stimuli are edited using Praat. S2-3. Use MATLAB to control the edited auditory stimuli; S2-4. Use Psychphysics Toolbox 3.0 to present auditory stimuli and complete fMRI data acquisition.

4. The auditory representation connectivity analysis method based on spatiotemporal information according to claim 1, characterized in that, In step S4, the method for analyzing brain region activation levels in task-oriented fMRI data to obtain the activated brain regions, which are the regions of interest (ROIs) for subsequent analysis, includes: S4-1. The general linear model (GLM) was used to analyze the brain region activation levels in task-state fMRI data. The calculation formula is as follows: Y = βX + ε, where Y represents the BOLD signal on each voxel, X represents some unknowns, β represents the linear combination, and ε represents the residual; S4-2. Use the least squares method to minimize the sum of squares of the residuals ε to fit the parameter β value; S4-3. Perform statistical inference on the β value and set a predetermined threshold to obtain the brain function activation map corresponding to the sound stimulus, which is the region of interest (ROI) for subsequent analysis.

5. The auditory representation connectivity analysis method based on spatiotemporal information according to claim 1, characterized in that, In step S6, the method for playing auditory stimuli and completing OPM-MEG data acquisition includes: S6-1. Put the helmet containing the OPM-MEG sensor on the subject and adjust it to a suitable position; S6-2, Play auditory stimuli and collect data.

6. The auditory representation connectivity analysis method based on spatiotemporal information according to claim 1, characterized in that, In step S7, the method for preprocessing OPM-MEG data to obtain a basic OPM-MEG signal and performing time-frequency analysis, calculating the average Z_score waveform, and taking the maximum value of the waveform as the response intensity includes: S7-1, Perform 1-30Hz offline bandpass filtering and plot the power spectral density to check and remove bad channels; S7-2. Perform data segmentation based on the recorded stimulus synchronization trigger signal, and extract data from 100ms before the stimulus and 1000ms after the stimulus; S7-3. Perform visual inspection to identify and remove abnormal segmented data obtained in S7-2, and obtain the basic OPM-MEG signal; S7-4. Use the Morlet wavelet-based continuous wavelet transform (CWT) to generate the time-frequency distribution of the basic OPM-MEG signal and obtain the corresponding time-frequency diagram. S7-5. Normalize the time-frequency graph and calculate the Z_score for each time point to obtain the Z_score waveform. Define the maximum value of the Z_score waveform as the intensity of the subject's response to the sound stimulus.

7. The auditory representation connectivity analysis method based on spatiotemporal information according to claim 1, characterized in that, In S9, the method for comparing and statistically testing the results obtained by the subjects in S7 and S8 to detect whether there are differences in the neural representation processes among the subjects includes: S9-1. Use the U statistic as the statistic for the Wilcoxon rank-sum test; S9-2. Define the null hypothesis as follows: The statistic of the intensity of the subject's response is not different from that of others. S9-3. Calculate the U statistic of the response intensity at each time point and sort the values ​​of the U statistic in order of magnitude; S9-4. Assign a rank R to each value; S9-5. Sum the ranks corresponding to each value to obtain the total rank sums U1 and U2. Based on the total rank sums U1 and U2, the U statistic is finally obtained. S9-6. Select P=0.05 as the significance level, and find the corresponding critical value of rank sum according to the Wilcoxon rank sum test distribution table; S9-7. If the rank sum obtained in S9-5 is greater than the critical value of the rank sum in S9-6, then reject the null hypothesis, that is, consider that the response strength is different from others; otherwise, accept the null hypothesis, that is, consider that the response strength is not different from others. S9-8: Based on the comparisons among subjects in S7 and the statistical test results in S9, we obtain the specific details of the subjects' auditory neural representation; based on the comparisons among subjects in S8 and the statistical test results in S9, we obtain the information flow and corresponding brain regions in the process of the subjects' auditory neural representation.

Images