Individualized rTMS precise intervention strategy based on chronic insomnia

Abstract

The invention discloses an individualized rTMS precise intervention strategy based on chronic insomnia, and relates to the technical field of nerve regulation and control. According to the method, individual differences are fully considered, and the neural physiological features of each patient are accurately identified by integrating fMRI-EEG multi-mode technologies, so that the rTMS intervention target spot is optimized. The personalized strategy can effectively reduce the heterogeneity of treatment, improve the intervention effect and provide a new solution for precise medical treatment of chronic insomnia.

Description

Technical Field

[0001] The present invention relates to the technical field of neuromodulation, and particularly relates to an individualized rTMS precise intervention strategy for chronic insomnia. Background Art

[0002] At present, many technologies can be applied to the intervention of chronic insomnia. Among them, repeated Transcranial Magnetic Stimulation (rTMS), as a non-invasive, safe and effective treatment method, has been widely used in the intervention of mental and neurological diseases. In recent years, rTMS has achieved remarkable research results in improving sleep quality and alleviating sleep disorder symptoms (Oroz et al., 2021). By continuously stimulating specific regions of the brain, such as the Dorsolateral Prefrontal Cortex (DLPFC), Superior Parietal Cortex (SPC), Ventromedial Prefrontal Cortex (vmPFC), and Visual Association Cortex (VAC), rTMS can not only effectively improve sleep quality, but also significantly enhance the cognitive decline caused by sleep deprivation, especially in working memory and spatial memory (Zhu et al., 2023). These findings provide strong support for rTMS as a treatment for cognitive problems caused by sleep disorders, and also provide a safe and effective adjuvant treatment method for coping with cognitive dysfunction caused by sleep disorders.

[0003] At the current stage, in the rTMS treatment of chronic insomnia, the mainstream method still relies on conventional positioning strategies, usually taking the Dorsolateral Prefrontal Cortex (DLPFC) as the main stimulation area (Rolls et al., 2023). However, this method has obvious deficiencies. Especially when dealing with individual differences, the treatment effects show significant heterogeneity (Drysdale et al., 2017). This is because the brain functional networks of chronic insomnia patients have obvious individual differences. At the same time, chronic sleep not only damages the neural representation of SWM, but also potentially disrupts the deep brain network mechanism (Palmer & Alfano, 2017), and the tolerance of different loads of SWM to sleep disorders is different (Zhang et al., 2025). These changes make the efficacy of traditional DLPFC-based interventions heterogeneous, difficult to ensure the stability of the efficacy among patients, and unable to fully meet the personalized needs of different patients.

[0004] During the rTMS intervention process, existing technologies usually rely on a neuronavigation system to ensure the accuracy of the stimulation target (Zheng et al., 2020), and the position of the stimulation coil is tracked and adjusted in real time through an infrared camera. However, this strategy mainly relies on the patient's structural MRI data to display the stimulation target, lacking real-time monitoring of the brain functional state during the rTMS treatment process and dynamic visualization of the target. This makes it difficult to implement an intervention mode based on real-time assessment of the patient's brain function and adjustment of rTMS treatment parameters, thus unable to ensure the effect of rTMS intervention on improving brain function.

[0005] Therefore, we propose an individualized rTMS precise intervention strategy for chronic insomnia. By implementing standardized information processing and multimodal fusion technology, an rTMS intervention plan for precise targets is formulated to provide personalized treatment decision support for chronic insomnia patients. Summary of the Invention

[0006] The purpose of the present invention is to provide an individualized rTMS precise intervention strategy for chronic insomnia to solve the above problems.

[0007] To achieve the above purpose, the technical solution adopted by the present invention includes the following steps: S1: Activate the working memory process of the chronic insomnia population using a spatial working memory task; In the spatial working memory task, a series of stimuli related to working memory are presented to the patient, and they are required to complete specific tasks; S2: Conduct the spatial working memory task and synchronously collect fMRI and EEG data. Use SPM 12 to preprocess the fMRI data and EEGLAB to preprocess the EEG data; S3: Use SPM 12 to register the preprocessed fMRI data and EEG data onto the MNI standard brain to ensure that the fMRI and EEG data are in the same coordinate system. Use the ROI_MNI_v4 version of the AAL atlas to map the fMRI and EEG data registered onto the MNI standard brain to the same brain regions and spatial coordinate system, and merge the fMRI and EEG data to obtain fused fMRI-EEG data; S4: Input the fused fMRI-EEG data into a generalized linear model, and adopt a functional connectivity DCM analysis method to identify the points with the strongest functional connectivity between the DLPFC, SPC, vmPFC, VAC, and the hippocampus in each patient during the spatial working memory task as the targets for individualized rTMS intervention. The specific content is as follows: Step1: Select the SPM.mat files generated by constructing the GLM model for all patients; Step 2: Select the VOI time series of the left DLPFC, right DLPFC, left PPC, right PPC, left vmPFC, right vmPFC, left VAC, and right VAC extracted from all patients; Step 3: Specify matrix A. This patent presupposes that there is an inherent balance mechanism for each VOI brain region and between brain regions; Step 4: Specify matrix B. This patent presupposes that the connections between all VOI brain regions are modulated by experimental tasks; Step 5: Specify matrix C. This patent presupposes that the inherent balance mechanisms of all VOI brain regions are modulated by experimental tasks; Step 6: Perform model estimation. Select the brain region with the strongest connection to the hippocampus as the rTMS intervention target; S5: Implement individualized rTMS intervention for the chronic insomnia population for 2 weeks; During the rTMS intervention process, combine an online EEG analysis algorithm to perform online denoising on the collected EEG data, including time-domain filtering, CSP filtering, online PCA, and online ICA; S6: Evaluate the clinical behavioral performance and cognitive assessment results of the patients before and after the intervention, and collect fMRI-EEG data.

[0008] Furthermore, use the 10-20 international standard lead system to set 64 sampling electrodes to obtain EEG data for the spatial working memory task, with a sampling frequency of 1000 Hz. The fMRI parameters are set as follows: voxel size is 3×3×3 mm³, matrix shape is 90×90, number of layers is 60, a total of 180 time points are collected, and the scanning duration is 6 minutes.

[0009] Furthermore, the specific content of the preprocessing of the fMRI and EEG data in S2 is as follows: The preprocessing of the pre-test EEG data using EEGLAB specifically includes: electrode positioning, deleting useless electrodes, 0.5-30 Hz band-pass filtering, downsampling rate to 250 Hz, data segmentation and baseline correction, excluding bad segments, performing ICA, removing eye electricity, electrocardiogram, and channel noise artifacts; The preprocessing of the pre-test fMRI data using SPM12 specifically includes: Step 1: Use MRIQC 23.2.0 to check the data and exclude motion artifacts and fluctuations; Step 2: Use SPM 12 for temporal slice correction; Step 3: Use SPM 12 for head motion correction. Adopt rigid motion correction to perform motion correction on the images of all time points. For the images of the time points , is the reference image, and the translation parameter is , and the rotation parameter is . The corrected image can be obtained through the following formula: (Formula 1) Step4: Map the fMRI data to the MNI standard brain; Step4: Use spatial smoothing to improve the signal-to-noise ratio, and the size of the smoothing kernel is 6 - 8 mm; use a 0.01 Hz high-pass filter to remove trends and low-frequency drifts in the data; Step5: Regress the fMRI data using the 6 rigid parameters of motion correction to remove motion-related signals. The process is as follows: (Formula 2) Where: is the fMRI data at time point , is the design matrix containing motion parameters, is the coefficient of the motion parameter, is the residual; Step6: Remove cerebrospinal fluid and white matter signals; And use SPM 12 to register the fMRI data and EEG data to the MNI standard brain uniformly to ensure that the fMRI and EEG data are in the same coordinate system; Then use the AAL atlas ROI_MNI_v4 version to map the fMRI and EEG data to the same brain regions and spatial coordinate system, and merge the fMRI and EEG data.

[0010] Furthermore, define a generalized linear model, use the binary-encoded spatial working memory task stimulus as the independent variable to construct a design matrix, use the physiological variable of the hippocampus when performing the spatial working memory task as the dependent variable, and use the least squares method to estimate the coefficient of the influence of the independent variable on the dependent variable. The calculation formula is as follows: (Formula 3) By adjusting the value of the coefficient vector , make the dependent variable best fit the linear combination of the design matrix , that is, minimize the error term; Select the left DLPFC, right DLPFC, left PPC, right PPC, left vmPFC, right vmPFC, left VAC, and right VAC as the VOIs.

[0011] Furthermore, the specific details of the time-domain filtering, CSP filtering, online PCA, and online ICA are as follows: For time-domain filtering, the high-pass filter is set to 0.1 Hz and the low-pass filter is set to 30 Hz; When using CSP filtering, different covariance matrices are created for different categories of EEG signals. The matrices are as follows: (Formula 4) Where, represents the variance of this factor and is calculated using the following formula:

[0012] (Formula 5) Where, is the EEG signal of different categories, is the mean vector of the data; represents the covariance of two factors and is calculated using the following formula: (Formula 6) Where, is the EEG signal of different categories, is the mean vector of the data; Perform a generalized eigenvalue decomposition on these covariance matrices. The basic idea is as follows: For a set of matrices , ... and , where, is a dimensional matrix, is a dimensional positive definite matrix. The form of the generalized eigenvalue problem is: (Formula 7) Where is a non-zero eigenvector, is the generalized eigenvalue. The generalized eigenvalue decomposition in the general form can be expressed using the following formula: (Formula 8) Where, is a dimensional matrix, is a dimensional matrix whose columns are combinations of the eigenvectors , is a diagonal matrix whose diagonal elements are the generalized eigenvalues , finally, the generalized eigenvalue decomposition will generate a set of eigenvalues and corresponding eigenvectors. Select the eigenvectors with the largest or smallest eigenvalues to construct filters. Select the eigenvectors with the largest eigenvalues to enhance the brain activities of interest, and select the eigenvectors with the smallest eigenvalues to suppress other brain activities. Finally, a set of CSP filters is generated, and each filter corresponds to a different brain activity pattern, some of which are used for enhancement and some for suppression; Online PCA analysis is performed on the synchronously collected EEG data, and the basic idea is as follows: When collecting EEG signals, there are n samplings, and each sampling has m electrodes. The EEG data is represented as a matrix , and the covariance matrix is calculated as follows: (Formula 9) where is the mean vector of the data. The covariance matrix is decomposed into eigenvalues using the following formula: (Formula 10) where is the eigenvector matrix, is a diagonal matrix, and the elements on the diagonal are eigenvalues. Select the eigenvectors corresponding to the first k eigenvalues to form the projection matrix , and project the data into the principal component space: (Formula 11) where is the data matrix after dimensionality reduction; The basic steps of online ICA are as follows: First, construct an EEG data matrix containing m electrodes, with a size of , where is the number of sampling points. ICA is mathematically expressed by the following formula: (Formula 12) where is the observed signal matrix, is the unknown mixing matrix, is the source signal matrix. The purpose of ICA is to find the source signal matrix , which specifically includes the following steps: Step1: For the data matrix , first perform centering using the following formula: (Formula 13) where is the mean of each column feature in the original data matrix ; Step 2: Whiten the data using the following formula: (Formula 14) Use the following formula to perform eigen - decomposition on the covariance matrix : (Formula 15) where is the eigen - vector matrix, is the diagonal eigenvalue matrix, and the whitening transformation can be expressed as the following formula: (Formula 16) Step 3: Based on maximizing non - Gaussianity, select negentropy as the measure of non - Gaussianity, use the non - linear function to approximate the maximization of negentropy, and find the projection weights that maximize negentropy. Its iterative update rule can be expressed by the following formula: (Formula 17) After that, use the following formula to normalize the new weight vector : (Formula 18) Step 4: Orthogonalization and convergence. Adopt the Gram - Schmidt process. For the th component, the update rule will consider the first components: (Formula 19) After normalizing , check for convergence; Step 5: Independent component analysis, find all weight vectors: (Formula 20) Estimate the independent components in the following way: (Formula 21).

[0013] Furthermore, for the EEG data after online denoising, use the source localization analysis model to perform weighted minimum - norm estimation inversion and solve the source analysis of the EEG data, which includes the following steps: Step 1: Establish a head model, including the following steps: Check the quality of the 3D T1 - weighted sequence images of the MRI to ensure that the images have no obvious motion artifacts, artifacts, or other artifacts; Register the MRI images with the spatial information of the electrode positions; Segment the MRI images using FreeSurfer; Use the 3D reconstruction software BrainSuite to convert the segmentation results of the MRI images into a 3D surface model; Step 2: Calculate the transfer matrix of the electrical signal on the cerebral cortex using the finite element method or the analytical method. The formula is as follows: (Formula 22) is the electrical conductivity of the tissue, and are the spatial positions of the source i and the electrode j respectively, represents the surface of the cerebral cortex, is the surface normal vector.

[0014] Step 3: Use the following formula for weighted minimum norm estimation: (Formula 23) where, is the estimate of the source signal, is the transfer matrix, is the observed EEG data, is the weighting matrix, is the regularization parameter; Step 4: Use the transfer matrix to convert the estimated source signal into the source activity distribution on the cerebral cortex. The formula is as follows: (Formula 24) where is the source signal, is the estimate of the electrode signal, is the pseudo-inverse matrix used to convert the electrode signal back to the source space; Step 5: Project the source activity onto the cerebral cortex surface of the head model, use Brainstorm to visualize the source activity on the cerebral cortex, overlay the pre-determined rTMS target point positions on the cerebral cortex, and compare them with the activated brain regions.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The individualized rTMS precise intervention plan designed by the present invention carries out refined regulation on the spatial working memory function of chronic insomnia patients, and has significant clinical and scientific research value. Compared with the traditional rTMS treatment method, the present invention fully considers individual differences, accurately identifies the neurophysiological characteristics of each patient by integrating fMRI-EEG multimodal technology, and then optimizes the rTMS intervention target points. This personalized strategy can effectively reduce the heterogeneity of treatment, improve the intervention effect, and provide a new solution for the precision medicine of chronic insomnia.

[0016] Technically, the present invention breaks through the dependence on fixed stimulation targets in traditional rTMS treatment, innovatively combines spatial working memory tasks, and dynamically monitors changes in key biomarkers during the treatment process. This method can identify brain network abnormalities and electrophysiological characteristics related to insomnia, thereby optimizing stimulation parameters and achieving the formulation of precise assessment and personalized treatment strategies. Through this scientific and individualized intervention method, the present invention can not only improve the treatment effect of rTMS on chronic insomnia, but also promote the research on the neural mechanisms related to spatial working memory, laying a foundation for the future development of neuromodulation technologies.

[0017] From a macroscopic perspective, the implementation of the present invention is expected to break through the limitations of current rTMS intervention programs and provide a new direction for the treatment of chronic insomnia. With the help of standardized information processing and multimodal fusion technologies, this program can accurately locate treatment targets and formulate individualized intervention strategies, thereby improving the stability and effectiveness of treatment. In addition, this method also provides a solid theoretical basis for the clinical application of precision medicine, promotes the wide application of neuromodulation technologies in the treatment of mental diseases, and brings a more scientific and effective individualized treatment plan for chronic insomnia patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is the flow chart of the rTMS precise intervention method strategy based on chronic insomnia of the present invention; Figure 2 is the task flow chart of the spatial working memory of the present invention; Figure 3 is the 64 sampling electrode diagram set by the 10-20 international standard lead system of the present invention; Figure 4 is the pre-experiment rTMS intervention efficacy diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] The following further describes the specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] As Figures 1 to 4 shown, this embodiment proposes an individualized rTMS precise intervention strategy based on chronic insomnia, including the following steps: As shown in Figure 2, patients with chronic insomnia performed a spatial working memory task, namely the pre-test. Three load levels of 0-back spatial working memory task, 1-back spatial working memory task, and 2-back spatial working memory task were used. The stimuli were located at four positions: up, down, left, and right. The stimulus presentation time and fixation point presentation time for each task were 400 ms and 1600 ms respectively. It was divided into two conditions: matching and non-matching with the target stimulus. Press 1 when matching, and do not press the button when non-matching. Each task had 180 trials, with a duration of 6 minutes, and the three tasks had a total duration of 18 minutes.

[0021] During the performance of the spatial working memory task, fMRI and EEG data were collected synchronously. SPM 12 was used to preprocess the fMRI data, and EEGLAB was used to preprocess the EEG data. The Pittsburgh Sleep Quality Index, Self-Rating Anxiety Scale, and Self-Rating Depression Scale were used to measure sleep status, and the severity of depressive and anxiety symptoms respectively.

[0022] As shown in Figure 3, 64 sampling electrodes were set using the 10-20 international standard lead system to obtain EEG data of the spatial working memory task, with a sampling frequency of 1000 Hz. At the same time, fMRI data of the spatial working memory task were collected, and the parameter settings were as follows: voxel size was 3×3×3 mm³, matrix shape was 90×90, the number of layers was 60, a total of 180 time points were collected, the scanning duration was 6 minutes. At the same time, a T1-weighted fast spoiled gradient echo sequence was used to obtain a structural image with the following scanning parameters: TR = 6.7 ms, TE = 2.7 ms, FOV = 224×100 mm 2 , slice thickness = 1 mm, flip angle = 8◦, matrix = 224×224 and 196 slices. The whole-brain scanning time was 6 minutes.

[0023] EEGLAB was used to preprocess the pre-test EEG data, including: electrode positioning, deleting useless electrodes, 0.5 - 30 Hz band-pass filtering, downsampling rate to 250 Hz, data segmentation and baseline correction, excluding bad segments, performing ICA, removing eye electricity, electrocardiogram, and channel noise artifacts.

[0024] SPM12 was used to preprocess the pre-test fMRI data, specifically including: Step1: Use MRIQC 23.2.0 to check the data and exclude motion artifacts and fluctuations.

[0025] Step2: Use SPM 12 for temporal slice correction.

[0026] Step 3: Use SPM 12 for head motion correction. Apply rigid motion correction to perform motion correction on the images at all time points. For the image at time point the image , is used as the reference image, the translation parameter is , and the rotation parameter is . The corrected image can be obtained through the following formula: (Formula 1) Step 4: Map the fMRI data to the MNI standard brain.

[0027] Step 4: Use spatial smoothing to improve the signal-to-noise ratio. The size of the smoothing kernel is 6 - 8 mm; use a 0.01 Hz high-pass filter to remove trends and low-frequency drifts in the data.

[0028] Step 5: Regress the fMRI data using the 6 rigid parameters of motion correction to remove motion-related signals. This process can be expressed by the following formula: (Formula 2) where: is the fMRI data at time point . is the design matrix containing motion parameters. are the coefficients of the motion parameters. is the residual.

[0029] Step 6: Remove cerebrospinal fluid and white matter signals.

[0030] Use SPM 12 to register the preprocessed fMRI data and EEG data to the MNI standard brain to ensure that the fMRI and EEG data are in the same coordinate system.

[0031] Use the AAL atlas ROI_MNI_v4 version to map the fMRI and EEG data registered to the MNI standard brain to the same brain regions and spatial coordinate system, and merge the fMRI and EEG data.

[0032] Define a generalized linear model, input the merged fMRI and EEG data, and use the binary-encoded spatial working memory task stimuli (e.g., 0 represents the target direction and 1 represents the non-target direction) as independent variables (psychological variables) to construct a design matrix. When performing the spatial working memory task, the physiological variables of the hippocampus are used as dependent variables, and the least squares method (Least Squares Estimation) is used to estimate the coefficients of the influence of the independent variables on the dependent variables. The calculation formula is as follows: (Formula 3) By adjusting the coefficient vector value, the dependent variable is best fitted to the linear combination of the design matrix , that is, minimizing the error term .

[0033] Select the left DLPFC, right DLPFC, left PPC, right PPC, left vmPFC, right vmPFC, left VAC, and right VAC as the VOIs.

[0034] Use SPM12 to construct a full DCM model and perform DCM analysis, including the following steps: Step1: Select the SPM.mat file generated by constructing the GLM model for all patients.

[0035] Step2: Select the VOI time series of the left DLPFC, right DLPFC, left PPC, right PPC, left vmPFC, right vmPFC, left VAC, and right VAC extracted from all patients.

[0036] Step3: Specify the A matrix. This patent presupposes that each VOI brain region and between brain regions has its own balance mechanism.

[0037] Step4: Specify the B matrix. This patent presupposes that the connections between all VOI brain regions are modulated by the experimental task.

[0038] Step5: Specify the C matrix. This patent presupposes that the balance mechanisms of all VOI brain regions themselves are modulated by the experimental task.

[0039] Step6: Perform model estimation, and select the brain region with the strongest connection to the hippocampus as the rTMS intervention target.

[0040] Individualized rTMS intervention lasts for 2 weeks, the stimulation intensity is 100% of the resting motor threshold (rMT), the stimulation frequency is 10 Hz, the stimulation duration is 1 s, and the interval is 10 s. The intervention lasts for 2 weeks, with continuous intervention for 5 working days per week, about 20 min per day.

[0041] During the rTMS intervention, combined with the online EEG analysis algorithm, online denoising of the collected EEG data is performed, including time-domain filtering, CSP filtering, online PCA, and online ICA.

[0042] For time-domain filtering, set the high-pass filter to 0.1 Hz and the low-pass filter to 30 Hz.

[0043] When using CSP filtering, different covariance matrices are created for different categories of EEG signals. The matrices are as follows: (Formula 4) Among them, represents the variance of this factor and is calculated using the following formula:

[0044] (Formula 5) Among them, is the EEG signal of different categories, is the mean vector of the data.

[0045] Represents the covariance of two factors and is calculated using the following formula: (Formula 6) Among them, is the EEG signal of different categories, is the mean vector of the data.

[0046] Generalized eigenvalue decomposition is performed on these covariance matrices. The basic idea is as follows: For a set of matrices , ... and , among them, is a - dimensional matrix, is a positive - definite matrix. The form of the generalized eigenvalue problem is: (Formula 7) Among them is a non - zero eigenvector, is the generalized eigenvalue. The generalized eigenvalue decomposition in this general form can be expressed by the following formula: (Formula 8) Among them, is a - dimensional matrix, is a - dimensional matrix whose columns are combinations of the eigenvectors . is a diagonal matrix whose diagonal elements are the generalized eigenvalues . Finally, the generalized eigenvalue decomposition will generate a set of eigenvalues and corresponding eigenvectors. Select the eigenvectors with the largest or smallest eigenvalues to construct the filter. Select the eigenvectors with the largest eigenvalues to enhance the brain activities of interest and select the eigenvectors with the smallest eigenvalues to suppress other brain activities. Finally, a set of CSP filters is generated, each filter corresponding to a different brain activity pattern, some of which are used for enhancement and some for suppression.

[0047] Online PCA analysis is performed on the synchronously collected EEG data, and the basic idea is as follows: When collecting EEG signals, there are n samplings, and each sampling has m electrodes (in this patent, the number of electrodes is 64). The EEG data is represented as a matrix , and the calculation formula of the covariance matrix is as follows: (Formula 9) where is the mean vector of the data. The covariance matrix is decomposed into eigenvalues using the following formula: (Formula 10) where is the eigenvector matrix, is the diagonal matrix, and the elements on the diagonal are the eigenvalues. Select the eigenvectors corresponding to the first k eigenvalues to form the projection matrix , and project the data into the principal component space: (Formula 11) where is the data matrix after dimensionality reduction.

[0048] The basic steps of online ICA are as follows: First, construct an EEG data matrix containing m electrodes, with a size of , where is the number of sampling points. ICA is expressed by the following formula in mathematics: (Formula 12) where is the observed signal matrix, is the unknown mixing matrix, is the source signal matrix, and the purpose of ICA is to find the source signal matrix , which specifically includes the following steps: Step1: For the data matrix , first perform centering using the following formula: (Formula 13) where is the mean of each column feature in the original data matrix .

[0049] Step2: Whiten the data using the following formula: (Formula 14) Use the following formula for the covariance matrix Perform eigenvalue decomposition: (Equation 15) where is the eigenvector matrix, is the diagonal eigenvalue matrix. The whitening transformation can be expressed as the following formula: (Equation 16) Step3: Based on maximizing non-Gaussianity, select negentropy as the measure of non-Gaussianity, and use the non-linear function to approximate the maximization of negentropy and find the projection weights that maximize negentropy. Its iterative update rule can be expressed by the following formula: (Equation 17) After that, use the following formula to normalize the new weight vector : (Equation 18) Step4: Orthogonalization and convergence, adopt the Gram-Schmidt process. For the th component, the update rule will consider the first components: (Equation 19) After normalizing , check for convergence.

[0050] Step5: Independent component analysis, find all weight vectors: (Equation 20) Estimate the independent components in the following way: (Equation 21) Use the source localization analysis model to perform weighted minimum norm estimation inversion to solve the source analysis of EEG data, including the following steps: Step1: Establish a head model, including the following steps: Check the quality of the 3D T1-weighted sequence images of the MRI to ensure that the images have no obvious motion artifacts, artifacts, or other artifacts; Register the MRI images with the spatial information of the electrode positions; Segment the MRI images using FreeSurfer; Use the 3D reconstruction software BrainSuite to convert the segmentation results of the MRI images into a 3D surface model.

[0051] Step2: Use the finite element method or the analytical method to calculate the transfer matrix of the electrical signal on the cerebral cortex. The formula is as follows: (Equation 22) is the conductivity of the tissue, and are the spatial positions of source i and electrode j respectively, represents the surface of the cerebral cortex, is the surface normal vector.

[0052] Step3: Perform weighted minimum norm estimation using the following formula: (Formula 23) where, is the estimate of the source signal, is the transfer matrix, is the observed EEG data, is the weighting matrix, is the regularization parameter.

[0053] Step4: Convert the estimated source signal to the source activity distribution on the cerebral cortex using the transfer matrix. The formula is as follows: (Formula 24) where is the source signal, is the estimated value of the electrode signal, is the pseudo-inverse matrix used to convert the electrode signal back to the source space.

[0054] Step5: Project the source activity onto the cerebral cortex surface of the head model, visualize the source activity on the cerebral cortex using Brainstorm, overlay the pre-determined rTMS target positions on the cerebral cortex, and compare with the activated brain regions.

[0055] Manually adjust the coil to ensure that the activated brain region and the intervention target are consistent.

[0056] Evaluate the efficacy of the patient after rTMS intervention.

[0057] Perform the working memory task again as the post-test, and simultaneously collect EEG and fMRI data.

[0058] Preprocess the pre-test EEG data using EEGLAB, including: electrode localization, deleting useless electrodes, 0.5 - 30 Hz band-pass filtering, downsampling to 250 Hz, data segmentation and baseline correction, removing bad segments, performing ICA, removing eye, heart, and channel noise artifacts.

[0059] Preprocess the post-test fMRI data using SPM12, with the same processing method as the pre-test.

[0060] When evaluating the efficacy of patients after rTMS intervention, the pre-test and post-test data after preprocessing and combination are analyzed using a general linear model (GLM), registered to a standard brain template for within-group averaging and multiple comparison correction. Modeling is performed in the second-order analysis, and paired-sample t-tests are used to calculate the n-back effect and activation regions between the pre-test and post-test, respectively. All analyses are corrected using GRF with a threshold of P.

[0061] It is determined according to the activation levels of the DLPFC and SPC before and after the intervention. If the activation level increases significantly, it is effective. The intervention efficacy is as Figure 4 shown.

[0062] In this embodiment, the present application also provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor. The characteristic is that when the processor executes the computer program, it implements an individualized rTMS precise intervention strategy for the spatial working memory task of the chronic insomnia population. The above has described the embodiments of the present invention in detail with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, without departing from the principle and spirit of the present invention, various changes, modifications, substitutions, and variations of these embodiments still fall within the protection scope of the present invention.

[0063] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0064] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A personalized rTMS precision intervention method for spatial working memory tasks in people with chronic insomnia, characterized in that: The following steps are involved: S1: Using spatial working memory tasks to activate working memory processes in people with chronic insomnia; In the spatial working memory task, patients are presented with a series of stimuli related to working memory and asked to complete a specific task; S2: Perform spatial working memory tasks and simultaneously collect fMRI and EEG data. Use SPM 12 to preprocess fMRI data and EEGLAB to preprocess EEG data. S3: Use SPM 12 to uniformly register the preprocessed fMRI data and EEG data to the MNI standard brain to ensure that the fMRI and EEG data are in the same coordinate system. Use the ROI_MNI_v4 version of AAL atlas to map the fMRI and EEG data registered to the MNI standard brain to the same brain region and spatial coordinate system, merge the fMRI and EEG data, and obtain fused fMRI-EEG data. S4: The fused fMRI-EEG data were input into a generalized linear model, and the functional connectivity DCM analysis method was used to identify the points with the strongest functional connectivity between DLPFC, SPC, vmPFC, VAC and hippocampus in each patient during the spatial working memory task as the targets for individualized rTMS intervention. The specific contents are as follows: Step 1: Select all patients to construct the SPM.mat file generated by the GLM model; Step 2: Select the VOI time series of the left DLPFC, right DLPFC, left PPC, right PPC, left vmPFC, right vmPFC, left VAC, and right VAC extracted from all patients; Step 3: Specify the A matrix. This patent presupposes that each VOI brain region and the brain regions have their own balance mechanism. Step 4: Specify the B matrix. This patent assumes that the connections between all VOI brain regions are modulated by the experimental task. Step 5: Specify the C matrix. This patent assumes that the balance mechanism of all VOI brain regions is modulated by the experimental task. Step 6: Perform model estimation and select the brain area with the strongest connection to the hippocampus as the rTMS intervention target; S5: Implement a 2-week individualized rTMS intervention for people with chronic insomnia; During the rTMS intervention, the collected EEG data were denoised online in combination with online EEG analysis algorithms, including time domain filtering, CSP filtering, online PCA, and online ICA; S6: Evaluate the patients’ clinical behavioral performance and cognitive assessment results before and after the intervention, and collect fMRI-EEG data.

2. A personalized rTMS precision intervention method for spatial working memory tasks for people with chronic insomnia as claimed in claim 1, characterized in that: The 10-20 international standard lead system was used to set up 64 sampling electrodes to obtain EEG data of the spatial working memory task. The sampling frequency was 1000 Hz. The fMRI parameters were set as follows: the voxel size was 3×3×3 mm³, the matrix shape was 90×90, the number of layers was 60, a total of 180 time points were collected, and the scanning time was 6 minutes.

3. The personalized rTMS precision intervention method for spatial working memory tasks for chronic insomnia patients as claimed in claim 1, characterized in that: The specific contents of the preprocessing of fMRI and EEG data in S2 are as follows: The pre-test EEG data were pre-processed using EEGLAB, including: electrode positioning, deleting useless electrodes, 0.5-30Hz bandpass filtering, downsampling rate to 250Hz, data segmentation and baseline correction, removing bad segments, performing ICA, removing electrooculogram, electrocardiogram, and channel noise artifacts; SPM12 was used to preprocess the pre-test fMRI data, including: Step 1: Use MRIQC 23.2.0 to check the data and exclude motion artifacts and fluctuations; Step 2: Use SPM 12 to perform time layer correction; Step 3: Use SPM 12 to correct head motion and rigid motion correction to correct the images at all time points. Image , is the reference image, and the translation parameter is , the rotation parameter is , the corrected image It can be obtained by the following formula: (Formula 1) Step 4: Map fMRI data onto the MNI standard brain; Step 4: Use spatial smoothing to improve the signal-to-noise ratio, with a smoothing kernel size of 6-8 mm; use a 0.01 Hz high-pass filter to remove trends and low-frequency drifts in the data; Step 5: Use the 6 rigid parameters of motion correction to regress the fMRI data to remove motion-related signals. The process is as follows: (Formula 2) in: It's at the time fMRI data, is the design matrix containing the motion parameters, are the coefficients of the motion parameters, is the residual; Step 6: Remove cerebrospinal fluid and white matter signals; And the fMRI data and EEG data were uniformly registered to the MNI standard brain using SPM 12 to ensure that the fMRI and EEG data were in the same coordinate system; Then use AAL atlas ROI_MNI_v4 version to map fMRI and EEG data to the same brain region and spatial coordinate system, and merge fMRI and EEG data.

4. The personalized rTMS precision intervention method for spatial working memory tasks for chronic insomnia patients as claimed in claim 1, characterized in that: Define the generalized linear model, use the binary-coded spatial working memory task stimulus as the independent variable to construct the design matrix, and use the physiological variables of the hippocampus as the dependent variable when performing the spatial working memory task. Use the least squares method to estimate the coefficient of the influence of the independent variable on the dependent variable. The calculation formula is as follows: (Formula 3) By adjusting the coefficient vector , so that the dependent variable and Design Matrix The linear combination of is the best fit, that is, the error term is minimized; The left DLPFC, right DLPFC, left PPC, right PPC, left vmPFC, right vmPFC, left VAC, and right VAC were selected as VOIs.

5. The personalized rTMS precision intervention method for spatial working memory tasks for chronic insomnia patients as claimed in claim 1, characterized in that: The specific contents of the time domain filtering, CSP filtering, online PCA and online ICA are as follows: For time domain filtering, set the high-pass filter to 0.1 Hz and the low-pass filter to 30 Hz; When using CSP filtering, different covariance matrices are created for different categories of EEG signals. The matrices are as follows: (Formula 4) in, Represents the variance of the factor and is calculated using the following formula: (Formula 5) in, For different types of EEG signals, is the mean vector of the data; Represents the covariance of two factors and is calculated using the following formula: (Formula 6) in, For different types of EEG signals, is the mean vector of the data; The generalized eigenvalue decomposition is used for these covariance matrices. The basic idea is as follows: for a set of matrices , ... and ,in, yes dimensional matrix, yes -dimensional positive definite matrix; the generalized eigenvalue problem is of the form: (Formula 7) in is a non-zero eigenvector, is the generalized eigenvalue, and the generalized eigenvalue decomposition in general form can be expressed by the following formula: (Formula 8) in, yes dimensional matrix, yes -dimensional matrix whose columns are eigenvectors A combination of is a diagonal matrix whose diagonal elements are generalized eigenvalues ,Finally, the generalized eigenvalue decomposition will generate a set of eigenvalues ​​and corresponding eigenvectors. The eigenvectors with the largest or smallest eigenvalues ​​are selected to construct filters. The eigenvector with the largest eigenvalue is selected to enhance the brain activity of interest, and the eigenvector with the smallest eigenvalue is selected to inhibit other brain activities. Finally, a set of CSP filters is generated, each filter corresponds to a different brain activity pattern, some of which are used for enhancement and some for inhibition; The basic idea of ​​online PCA analysis of synchronously collected EEG data is as follows: when collecting EEG signals, there are n samplings, each sampling has m electrodes, and the EEG data is represented as a The matrix , the covariance matrix The calculation formula is as follows: (Formula 9) in, is the mean vector of the data, and the covariance matrix is ​​decomposed into eigenvalues ​​using the following formula: (Formula 10) in, is the eigenvector matrix, It is a diagonal matrix. The elements on the diagonal are eigenvalues. The eigenvectors corresponding to the first k eigenvalues ​​are selected to form the projection matrix , project the data into the principal component space: (Formula 11) in, is the data matrix after dimensionality reduction; The basic steps of online ICA are as follows: First, construct an EEG data matrix containing m electrodes , the size is ,in is the number of sampling points. ICA is expressed mathematically as follows: (Formula 12) in, is the observation signal matrix, is the unknown mixing matrix, is the source signal matrix. The purpose of ICA is to find the source signal matrix , specifically including the following steps: Step 1: For the data matrix , first use the following formula to decentralize: (Formula 13) in, is the original data matrix The mean of each column feature; Step 2: Whiten the data using the following formula: (Formula 14) The covariance matrix is ​​calculated using the following formula Perform eigendecomposition: (Formula 15) in, is the eigenvector matrix, It is a diagonal eigenvalue matrix. The whitening transformation can be expressed as follows: (Formula 16) Step 3: Based on maximizing non-Gaussianity, negative entropy is selected as the measure of non-Gaussianity, using a nonlinear function To approximate the maximization of negative entropy and find the projection weight that maximizes negative entropy, its iterative update rule can be expressed by the following formula: (Formula 17) Afterwards, the new weight vector is calculated using the following formula: Normalize: (Formula 18) Step 4: Orthogonalization and convergence, using the Gram-Schmidt process, for the components, the update rule will take into account the previous Quantity: (Formula 19) right After normalization, check whether it converges; Step 5: Independent component analysis, find all weight vectors: (Formula 20) The independent components are estimated as follows: (Official 21).

6. The personalized rTMS precision intervention method for spatial working memory tasks for chronic insomnia patients as claimed in claim 1, characterized in that: For the EEG data after online denoising, the source localization analysis model is used to perform weighted minimum norm estimation inversion solution and source tracing analysis on the EEG data, which includes the following steps: Step 1: Establish a head model, including the following steps: perform a quality check on the 3DT1-weighted sequence images of MRI to ensure that the images have no obvious motion artifacts, artifacts or other artifacts; align the MRI images with the spatial information of the electrode positions; use FreeSurfer to segment the MRI images; use the 3D reconstruction software BrainSuite to convert the segmentation results of the MRI images into a 3D surface model; Step 2: Use the finite element method or analytical method to calculate the transfer matrix of electrical signals on the cerebral cortex. The formula is as follows: (Formula 22) is the electrical conductivity of the tissue, and are the spatial positions of source i and electrode j, represents the surface of the cerebral cortex, is the surface normal vector; Step 3: Use the following formula to perform weighted minimum norm estimation: (Formula 23) in, is an estimate of the source signal, is the transfer matrix, is the observed EEG data, is the weighting matrix, is the regularization parameter; Step 4: Use the transfer matrix to convert the estimated source signal into the source activity distribution on the cerebral cortex. The formula is as follows: (Official 24) in is the source signal, is an estimate of the electrode signal, is the pseudo-inverse matrix used to transform the electrode signals back to the source space; Step 5: Project the source activity onto the cerebral cortex surface of the head model, use Brainstorm to visualize the source activity on the cerebral cortex, superimpose the predetermined rTMS target location on the cerebral cortex, and compare it with the activated brain area.

Images