Transcranial magnetic stimulation positioning system based on deep effect brain region

Abstract

The invention discloses a transcranial magnetic stimulation positioning system based on a deep effect brain region, and the system comprises the steps: obtaining the data of a tested magnetic resonance image, which comprises a resting state functional image and a high-resolution T1 structure image; performing data preprocessing on the function image and the structure image, and establishing an individual space; the deep effect brain region template is converted to an individual space through nonlinear registration, and a target region of interest is defined; calculating functional connections between voxels of the deep effect brain region and the cortical region by adopting a Granger causality analysis method based on wavelet transformation; and screening Top block masses according to the functional connection strength, calculating the gravity centers of the Top block masses, and determining individualized stimulation targets in combination with the skull surface distance. According to the method, spontaneous brain activity function variation of the brain of an individual and the brain signal propagation direction are fully considered, accurate positioning of different individual levels is achieved, relatively stable cortex stimulation targets are obtained, and therefore the clinical curative effect of TMS is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of biomedical image pattern recognition, and in particular to a transcranial magnetic stimulation positioning system based on deep effect brain regions. Background Art

[0002] In 2022, the U.S. Food and Drug Administration (FDA) approved Stanford Neuromodulation Therapy (SNT) for the treatment of refractory depression. Individualized rTMS therapy guided by functional magnetic resonance imaging (fMRI) has garnered widespread attention. Transcranial magnetic stimulation (TMS), a safe and non-invasive neuromodulation method, has been widely used in the treatment of various neuropsychiatric disorders. Research and analysis have found that TMS can effectively improve the symptoms of various neuropsychiatric disorders. However, improving the efficacy of TMS for various neuropsychiatric disorders remains a key challenge.

[0003] First, the target of transcranial magnetic stimulation (TMS) may be key to improving therapeutic efficacy. Traditionally, the location of the stimulation target has been based on conventional methods, which primarily involves using a "hotspot" in the hand area, moving it 5 cm forward to locate the left dorsolateral prefrontal cortex (DLPFC), or moving it a certain distance in other directions to locate other cortical regions, such as the supplementary motor area (SMA). This relatively crude location method ignores individual differences in cortical structure, and the stimulation target is not necessarily located in the target cortical region.

[0004] Furthermore, the modulatory effects of TMS are closely related to coil type, stimulation parameters, and coil targeting. The most common figure-8 TMS coil achieves excellent focusing, with a precision of approximately 3mm, but only on the cortical surface. Neuropsychiatric pathology may reside in deep brain regions, and effectively modulating these regions may be key to improving its efficacy. In fact, the modulatory effects of TMS are achieved by linking cortical stimulation targets with deep brain regions, thereby modulating brain activity in these and other regions. However, functional connectivity reliability has long been a concern, with traditional functional connectivity reliability being poor, as it directly determines the stability of the stimulation target. Improving stimulation target stability and effectively modulating brain activity in deep brain regions, as well as between deep brain regions and other regions, may be key to treating depression. Therefore, individualizing TMS based on functional connectivity holds important implications for future clinical research and application. Summary of the Invention

[0005] The purpose of this invention is to provide a transcranial magnetic stimulation (TMS) targeting system based on deep-effect brain regions. This system fully considers the individual variation in spontaneous brain activity and the direction of brain signal propagation, achieving precise targeting at different individual levels and obtaining relatively stable cortical stimulation targets, thereby optimizing the clinical efficacy of TMS.

[0006] The technical solution of the present invention is a transcranial magnetic stimulation positioning system based on deep effect brain areas, comprising:

[0007] Data acquisition module, used to obtain the subject's magnetic resonance imaging data, including resting-state functional images and high-resolution T1 structural images;

[0008] A preprocessing module, connected to the data acquisition module, is used to perform time point deletion, head motion correction, spatial normalization, noise removal, and wavelet filtering on the functional image and the structural image to establish an individual space;

[0009] A registration module, connected to the preprocessing module, converts the deep effect brain region template into the individual space through nonlinear registration to define the target region of interest;

[0010] a calculation module, connected to the registration module, and using a Granger causality analysis method based on wavelet transform to calculate the functional connectivity between the deep effect brain area and each voxel in the cortical area;

[0011] The target positioning module is connected to the calculation module, and screens the top clusters according to the functional connection strength, calculates their weighted geometric center of gravity, and determines the individualized stimulation target in combination with the distance from the skull surface.

[0012] The above-mentioned transcranial magnetic stimulation positioning system based on deep effect brain areas, the pre-processing module includes:

[0013] A time correction unit, used to delete the first n time points of the function image;

[0014] The spatial alignment unit uses linear affine transformation to align the functional image and the structural image;

[0015] Noise removal unit, using Friston-24 parameters to regress head motion, white matter and cerebrospinal fluid signals;

[0016] The filtering unit performs 0.01-0.08 Hz bandpass filtering through continuous wavelet transform, wherein the wavelet function uses Daubechies db2 basis function.

[0017] In the aforementioned transcranial magnetic stimulation positioning system based on deep effect brain areas, in the calculation module, the calculation formula of the continuous wavelet transform is as follows:

[0018]

[0019] Where: X t and Y t They refer to the time series of two voxels respectively, k refers to the translation parameter of the wavelet, ψ0 is the wavelet function; s refers to the translation parameter of the wavelet, and t refers to time.

[0020] In the aforementioned transcranial magnetic stimulation positioning system based on deep effect brain regions, in the registration module, the deep effect brain regions are defined by:

[0021] The inverse transformation matrix of the MNI152 standard space template was used to transform the preset coordinate spherical ROI into the individual space.

[0022] In the aforementioned transcranial magnetic stimulation positioning system based on deep effect brain areas, in the calculation module, the functional connectivity calculation adopts the following Granger causality model:

[0023]

[0024] Where: X t and Y t are the wavelet coefficients of the deep effect brain area and cortical voxels respectively; p represents the hysteresis coefficient; A k B is the historical influence weight of deep brain area X on cortex Y; k is the autoregressive weight of the historical state of cortex Y itself; A' k is the historical influence weight of cortex Y on deep brain area X, B' k is the autoregressive weight of the historical state of deep brain region X itself; C and C' are the regression coefficients of the covariates respectively; E t and E t ' are error terms; Z t is a covariate.

[0025] In the aforementioned transcranial magnetic stimulation positioning system based on deep effect brain areas, in the target positioning module, the top clusters are screened at intervals of 5%, and the voxels with the top 10%-50% Granger causality values ​​in the left dorsolateral prefrontal cortex are selected. The cluster boundary value is set to 26 and the number of voxels is greater than 20. The center of gravity of the cluster under each threshold is calculated as the candidate target.

[0026] In the aforementioned transcranial magnetic stimulation positioning system based on deep effect brain areas, the center of gravity is calculated as the weighted geometric center of gravity:

[0027]

[0028] Where: X i is the 3D coordinate of the voxel, n j is the total number of voxels contained in the jth cluster; w i Each voxel coordinate value is weighted by the Granger causality value of its voxel.

[0029] For the aforementioned transcranial magnetic stimulation positioning system based on deep effect brain areas, the final stimulation target must be located at the center of gravity of the top 20% functional connection cluster in the Ranger causality value and be less than 30 mm away from the skull surface.

[0030] The aforementioned transcranial magnetic stimulation positioning system based on deep effect brain areas,

[0031] The filtering unit adopts a multi-scale credibility optimization strategy, specifically including:

[0032] Frequency band selection unit: preferentially selects the 0.06-0.125 Hz frequency band for functional connectivity calculation;

[0033] Basis function verification unit: dynamically selects the optimal basis function by comparing the reliability index ICC ≥ 0.4 voxel number of Daubechies db2 / db4 / db6 basis functions;

[0034] Boundary compensation unit: The mirror extension method is used to eliminate the wavelet transform boundary effect on the edge voxels of individual space.

[0035] Compared with existing technologies, this invention significantly improves the precise positioning capability of transcranial magnetic stimulation (TMS) through multimodal data fusion and innovative signal processing technology. The specific beneficial effects are as follows:

[0036] 1. This method uses a continuous wavelet transform (CWT) combined with a db2 filter for 0.01-0.08Hz filtering. Compared to traditional FFT methods, this method improves functional connectivity reliability by a factor of 2, effectively reduces inter-scan noise interference, and ensures target stability. By systematically comparing the effects of different wavelet methods (MODWT / CWT), filters (Daubechies series), and lengths (2-20), this method selects the optimal parameter combination to maximize the repeatability of functional connectivity.

[0037] 2. This invention proposes using deep-effect brain regions (such as the pgACC) as seed points to indirectly regulate deep structures through cortical-deep functional connectivity loops, overcoming the limitation of traditional figure-8 coils that cannot directly stimulate deep brain regions. Based on Granger causality analysis (GCA), this invention clarifies the direction of signal transmission (e.g., DLPFC to pgACC), achieving targeted neuromodulation and enhancing the specificity of treatment.

[0038] 3. This method combines the distance between the center of gravity of the top 20% of clusters and the skull to ensure that the target is located in the stimulable area of ​​the cortex, avoiding ineffective deep stimulation. This method verifies the stability of the target through threshold analysis to reduce the impact of individual differences.

[0039] 4. This invention integrates resting-state functional images with high-definition T1 structural images, using spatial normalization and nonlinear registration techniques to precisely map individual brain structural differences and improve the accuracy of target anatomical localization. The preprocessing of this invention regresses white matter, cerebrospinal fluid, and head motion signals to eliminate non-neural signal interference and enhance the reliability of functional connectivity analysis.

[0040] In summary, the present invention fully considers the spontaneous functional variation of brain activity in an individual's brain and the direction of brain signal propagation, achieves precise positioning at different individual levels, and obtains relatively stable cortical stimulation targets, thereby optimizing the clinical efficacy of TMS. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a structural diagram of the system of the present invention;

[0042] Figure 2 This is a functional connectivity diagram of the individualized pgACC as the seed point and DLPFC in the embodiment;

[0043] Figure 3 Figure A is the functional connectivity diagram of Granger causality analysis in the x→y direction, Figure B is the functional connectivity diagram of Granger causality analysis in the y→x direction, and Figure C is the traditional functional connectivity diagram based on Pearson correlation;

[0044] Figure 4 Individualized cortical stimulation targets identified in the examples;

[0045] Figure 5 The individualized cortical stimulation targets finally determined in the embodiment;

[0046] Figure 6 It is a comparison of the local index stability (ICC) obtained based on different wavelet transforms, different mother wavelets, and different wavelet lengths with the traditional Fourier transform;

[0047] Figure 7 It is a schematic diagram of the operation flow of the system of the present invention;

[0048] Figure 8 are target location maps of two stimulation targets obtained by the method of the present invention, corresponding to map A and map B respectively;

[0049] Figure 9 The target position maps of the two stimulation targets obtained by the traditional peak point method correspond to Figure A and Figure B respectively; DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0051] Example: A transcranial magnetic stimulation positioning system based on deep effect brain areas, such as Figure 1 As shown, it consists of the following modules:

[0052] The data acquisition module 101 includes:

[0053] A 3T magnetic resonance scanner (Siemens MAGNETOM Prisma) was used with a resting-state fMRI sequence (TR = 2000 ms, TE = 30 ms, slice thickness 3.2 mm) and a 3D-T1 structural imaging sequence (slice thickness 1 mm).

[0054] Data storage server. Receives and stores original DICOM format image data in real time.

[0055] The pre-processing module 102 includes:

[0056] A computing workstation running the MATLAB environment, with SPM12 and the FSL toolkit built in, performs the following processing flow:

[0057] Time correction unit: automatic deletion function like the first 20 time points;

[0058] Spatial alignment unit: uses the FLIRT algorithm to perform affine registration of functional images and structural images;

[0059] Noise removal unit: removes head motion, white matter and cerebrospinal fluid signals through Friston-24 parameter regression;

[0060] Wavelet filtering unit: Call the Wavelab toolbox to perform continuous wavelet transform (db2 basis function, 0.01-0.08 Hz bandpass filtering).

[0061] Registration module 103: Using the ANTs nonlinear registration tool, the deep effect brain area in the MNI152 space template (a spherical ROI with a radius of 5 mm centered at coordinates [0, 42, 6]) was reversely mapped to the individual space to generate an individualized pgACC template.

[0062] The calculation module 104 includes:

[0063] GPU-accelerated server (NVIDIA Tesla V100), running customized Granger causality analysis algorithm:

[0064] Extract the wavelet coefficients of the pgACC seed point and the left DLPFC voxel;

[0065] The causal effect direction was calculated based on the bivariate autoregressive model, and a functional connectivity strength map in the DLPFC→pgACC direction was output.

[0066] The target location module 105 includes:

[0067] Threshold screening unit: extract the top 10%-50% voxels with high connectivity at an interval of 5%;

[0068] Cluster analysis unit: identifies connected regions with voxel number > 20 and boundary value 26;

[0069] Center of gravity calculation unit: performs weighted geometric center of gravity calculation on the top 20% of clusters:

[0070] Distance verification unit: calls 3D Slicer to calculate the distance between the center of gravity and the skull surface.

[0071] This embodiment also provides an operation process of the system including:

[0072] Data collection:

[0073] The subjects were placed supine on the MRI scanner bed with their heads secured with a foam pad. A 5-minute resting-state fMRI scan was first performed, followed by a high-definition T1 structural image.

[0074] Automated processing:

[0075] The preprocessing module completes data cleaning, registration, and wavelet filtering within 40 minutes to generate an individual spatial functional connectivity map.

[0076] Target identification

[0077] The system automatically screened out three candidate targets and selected the top 20% center of gravity 15.6 mm away from the skull as the final target.

[0078] Specifically, such as Figure 7 As shown, the operation process steps are as follows:

[0079] Step 1: The subject undergoes an MRI scan to obtain a resting-state functional image and a high-resolution T1 structural image of the subject. In this step, the scanning parameters of the MRI scan are as follows:

[0080] fMRI scanning parameters: repetition time (TR) = 2000ms; echo time (TE) = 30ms; flipangle (FA) = 90; 43 slices with interleaved acquisition; matrix = 64*64; field of view (FOV) = 220mm; acquisition voxel size = 3.44mm*3.44mm*3.20mm;

[0081] 3D-T1 scanning parameters: 176 sagittal slices, thickness = 1 mm, TR = 8.1 ms, TE = 3.1 ms, FA = 8, FOV = 250 mm.

[0082] Step 2: Data preprocessing of functional and structural images, including time point deletion, head motion correction, spatial normalization, noise removal, and wavelet filtering, to establish individual space;

[0083] The data preprocessing in step 2 specifically includes:

[0084] Step 2.1, delete the first 20 time points of the functional image;

[0085] Step 2.2: Align the functional image and the structural image using a linear affine transformation. In this step, the functional images at all time points are aligned to the structural image using a linear affine transformation, and the structural image is segmented and the brain tissue signal is extracted.

[0086] Step 2.3: Segment the structural image and extract brain tissue signals such as gray matter, white matter, and cerebrospinal fluid. Then, based on the individual level, regress out white matter, cerebrospinal fluid, head motion, and whole-brain means using a general linear model (Friston-24 parameters).

[0087] Step 2.4: Apply a 6mm full-width at half-maximum Gaussian kernel for spatial smoothing to reduce image noise;

[0088] Step 2.5: Perform 0.01-0.08 Hz bandpass filtering by continuous wavelet transform to eliminate noise signals, wherein the wavelet function uses Daubechies db2 basis function;

[0089] The calculation formula of the continuous wavelet transform is as follows:

[0090]

[0091] Where: X t and Y t They refer to the time series of two voxels respectively, k refers to the translation parameter of the wavelet, ψ0 is the wavelet function; s refers to the translation parameter of the wavelet, and t refers to time.

[0092] Step 3: Transform the deep-effect brain region template to individual space through nonlinear registration to define the target region of interest. In this step, the deep brain region (e.g., pgACC) in the standard template (MNI152) is precisely mapped to the subject's individual space to form an individualized region of interest (ROI). This step includes preliminary alignment of the functional and structural images to eliminate head motion and gross spatial differences. Using a pixel-by-pixel deformation field (e.g., Demons algorithm or B-spline transform), the individual structural image is locally adjusted to match the standard template to capture subtle differences such as cortical folds and brain volume. The inverse transformation matrix of the deep brain region ROI in the standard template is used to transform the pre-set coordinate spherical ROI (the deep-effect brain region is defined as the pgACC, defined as a spherical ROI centered on the MNI coordinates [0, 42, 6] with a radius of 5 mm) to individual space.

[0093] Step 4: Use the wavelet transform-based Granger causality analysis method to calculate the functional connectivity between the deep effect brain area and each voxel in the cortical area; in this step, the functional connectivity is the effect connection between the seed point and each voxel in the cortical area, that is, the wavelet transform-based Granger causality analysis of the average time series of the deep effect brain area and each voxel in the target cortical area.

[0094] Functional connectivity is based on the deep effect brain area of ​​wavelet transform, rather than the traditional functional connectivity based on Fourier transform. The so-called functional connectivity refers to the Granger causality (GCA) between wavelet coefficients based on wavelet transform:

[0095] Functional connectivity calculation uses the following Granger causality model:

[0096]

[0097] Where: X t and Y t are the wavelet coefficients of the deep effect brain area and cortical voxels respectively; p represents the hysteresis coefficient; A k B is the historical influence weight of deep brain area X on cortex Y; k is the autoregressive weight of the historical state of cortex Y itself; A' k is the historical influence weight of cortex Y on deep brain area X, B' k is the autoregressive weight of the historical state of deep brain region X itself; C and C' are the regression coefficients of the covariates respectively; E t and E t ' are error terms; Z t is a covariate.

[0098] In this embodiment, the cortical area is the left DLPFC, and the functional connectivity is the functional connectivity based on the pgACC as the seed point. With the pgACC as the seed point, the GCA of the wavelet coefficients obtained by wavelet transform of the pgACC and the wavelet coefficients of each voxel in the left DLPFC is calculated, i.e., the Granger causality. Figure 2 As shown, there is a shadow region (A). In addition, traditional functional connectivity only reflects the synchronization of brain area activities and cannot distinguish direction or causality. Granger causality provides directional information through the predictive relationship of time series, making up for the non-directional defect of traditional functional connectivity. Figure 3 As shown, Figure A is the functional connectivity diagram of Granger causality analysis in the x→y direction, Figure B is the functional connectivity diagram of Granger causality analysis in the y→x direction, and Figure C is the traditional functional connectivity diagram based on Pearson correlation.

[0099] Step 5: Filter the top clusters based on functional connectivity strength, calculate their centroids, and determine individual stimulation targets based on the distance from the skull surface. In this step, stimulation targets are located based on the centroids of the GCA-derived clusters within the top functional connectivity clusters.

[0100] The cluster centroid calculation method based on GCA results is as follows:

[0101] The method for locating the stimulation target is to use the center of gravity of the cluster based on the GCA result, and the calculation method is as follows:

[0102] First, we selected voxels with the top 10% to top 50% GCA values ​​in the DLPFC (i.e., GCA values ​​from voxels in the DLPFC to voxels in deep effector brain regions greater than 0) at intervals of 5%. We saved the cluster for each threshold and set the cluster boundary value to 26, i.e., the points were connected and the cluster size was greater than 20 voxels.

[0103] The center of gravity of the cluster under each threshold is taken as the potential stimulation target.

[0104] The centroid of Clusterj is calculated as:

[0105]

[0106] Where: X i is the 3D coordinate of the voxel, n j is the total number of voxels contained in the jth cluster; w i Each voxel coordinate value is weighted by the Granger causality value of its voxel.

[0107] Potential stimulation targets were statistically determined based on different thresholds. When the top 20% of stimulation targets were reached, the location of the stimulation target became stable and did not change. Finally, the center of gravity of the top 20% of clusters with supracortical effect connectivity, located within 30 mm of the skull surface, was selected as the final stimulation target.

[0108] In this embodiment, the individual spatial coordinates of the voxel with the strongest negative functional connectivity are: [-44, 77, 59], such as Figure 4 As shown; the negative functional connectivity strength is: 0.395; the cortical depth of this voxel position is: 15.6mm, which is less than 30mm, so the individualized cortical stimulation target is finally determined to be: [-34, 64, 54], as shown Figure 5 The intersection shown.

[0109] Furthermore, the present invention systematically evaluates the impact of different wavelet methods, wavelet filters, and wavelet lengths on functional connectivity reliability, and compares it with traditional fast Fourier transform (FFT) functional connectivity reliability. This systematically evaluates the impact of different factors on the stability of functional connectivity signals, that is, the impact on reliability between two scans. The specific steps are as follows:

[0110] 1. After completing the aforementioned data preprocessing, two wavelet transform methods, maximum overlap discrete wavelet transform (MODWT) and continuous wavelet transform (CWT), were used, and different wavelet filters (Daubechies Extreme Phase and Daubechies Least Asymmetric) and different wavelet lengths (2 to 20) were used for filtering, and the data were divided into different frequency bands (band 1: 0.125-0.25 Hz, band 2: 0.06-0.125 Hz, band 3: 0.03-0.06 Hz, band 4: 0.015-0.03 Hz);

[0111] 2. Calculate the functional connectivity under the above different situations (2 wavelet transform methods * 2 filters * different wavelet lengths [2 to 20]);

[0112] 3. Based on the functional connectivity results between the two scans, ICC was used for reliability analysis, and the number of voxels with ICC>=0.4 between the two scans, that is, with medium and high reliability, was calculated.

[0113] 4. Compare the number of voxels with ICC>=0.4 between the two scans under different conditions.

[0114] Reliability comparison Figure 6As shown in the figure, the results show that in most cases, the reliability of wavelet-based functional connectivity is better than that of traditional functional connectivity, especially at scale 2 (0.06-0.125Hz), the reliability of db2-FC under the CWT method (the number of voxels with ICC>=0.4) is twice that of traditional functional connectivity.

[0115] Furthermore, according to the reliability comparison result, the filtering unit adopts a multi-scale reliability optimization strategy, which specifically includes:

[0116] Frequency band selection unit: preferentially selects the 0.06-0.125 Hz frequency band for functional connectivity calculation;

[0117] Basis function verification unit: dynamically selects the optimal basis function by comparing the reliability index ICC ≥ 0.4 voxel number of Daubechies db2 / db4 / db6 basis functions;

[0118] Furthermore, the present invention uses a Granger Causality Analysis (GCA) method based on wavelet transform (db2 wavelet under the CWT method) to calculate the functional connectivity between deep effect brain regions and voxels in the cortical region. The top 20% clusters are selected based on the strength of the functional connectivity, their centers of gravity are calculated, and individualized stimulation targets are determined based on the distance from the skull surface. This is compared with the traditional fast Fourier transform (FFT) method that determines individualized stimulation targets based on peak points. The system evaluates the difference in coordinates of the stimulation targets determined between two scans of the same subject, that is, the impact of the stability of the stimulation targets between the two scans. The specific steps are as follows:

[0119] 1. After the aforementioned data preprocessing, the two scan data of the same subject were filtered in the 0.01-0.08 frequency band using the CWT wavelet transform method, db2 wavelet, and traditional fast Fourier transform (FFT).

[0120] 2. Calculate functional connectivity in the above different situations (Granger causality analysis method GCA and traditional Pearson linear correlation);

[0121] 3. Determine the target coordinates based on the functional connectivity results between the two scans using the center of gravity COG and traditional peak point methods;

[0122] 4. Compare the intra-individual Euclidean distance (ED) of the target coordinates of the two scans under different conditions.

[0123] Based on wavelet filtering, Granger causality analysis (GCA) was used to calculate the functional connectivity between the deep effect brain area and each voxel in the cortical area. The top 20% clusters were selected according to the functional connectivity strength, and their centers of gravity were calculated to determine the individual stimulation targets for the same subject in two scans, respectively [-38, 62, 56] and [-44, 68, 50], with an intra-individual Euclidean distance (ED) of 10.39 mm. The target positions on the subject's structural items are as follows: Figure 8 As shown in Figures A and B.

[0124] The traditional peak point method calculates the stimulation target points of the same subject in two scans, which are [-48, 63, 48] and [-57, 63, 36], respectively. The intra-individual Euclidean distance (ED) is 15 mm. The target position on the subject's structure is as follows: Figure 9 As shown in Figures A and B.

[0125] The results show that the target stability of the cluster center calculation scheme is better than that of the conventional scheme.

[0126] In summary, the present invention addresses the problem that traditional TMS positioning systems do not take into account the differences caused by individual brain lesions and the "8"-shaped coil cannot directly act on the deep brain. Based on resting-state magnetic resonance imaging data, the present invention uses a new effector brain area pgACC and a method of deep effector brain area effector connection. The strength of the effector connection, the stability of the stimulation target, and the distance between the voxel position and the skull are comprehensively evaluated to obtain an individualized DLPFC stimulation target. Through this functional connection loop, it indirectly acts on the deep effector brain area, thereby optimizing the antidepressant effect of TMS.

[0127] Furthermore, in this embodiment, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiment described above is merely illustrative. For example, the division of the units described is merely a logical functional division. In actual implementation, other division methods may be used. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not implemented. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units through some interface, which may be electrical or other forms.

[0128] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0129] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0130] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

Claims

1. A transcranial magnetic stimulation positioning system based on deep effect brain areas, characterized by: include: Data acquisition module, used to obtain the subject's magnetic resonance imaging data, including resting-state functional images and high-resolution T1 structural images; A preprocessing module, connected to the data acquisition module, is used to perform time point deletion, head motion correction, spatial normalization, noise removal, and wavelet filtering on the functional image and the structural image to establish an individual space; A registration module, connected to the preprocessing module, converts the deep effect brain region template into the individual space through nonlinear registration to define the target region of interest; a calculation module, connected to the registration module, and using a Granger causality analysis method based on wavelet transform to calculate the functional connectivity between the deep effect brain area and each voxel in the cortical area; The target positioning module is connected to the calculation module, and screens the top clusters according to the functional connection strength, calculates their weighted geometric center of gravity, and determines the individualized stimulation target in combination with the distance from the skull surface.

2. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 1, characterized in that: The pre-processing module comprises: A time correction unit, used to delete the first n time points of the function image; The spatial alignment unit uses linear affine transformation to align the functional image and the structural image; Noise removal unit, using Friston-24 parameters to regress head motion, white matter and cerebrospinal fluid signals; The filtering unit performs 0.01-0.08 Hz bandpass filtering through continuous wavelet transform, wherein the wavelet function uses Daubechies db2 basis function.

3. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 1, characterized in that: In the calculation module, the calculation formula of continuous wavelet transform is as follows: Where: X t and Y t They refer to the time series of two voxels respectively, k refers to the translation parameter of the wavelet, ψ0 is the wavelet function; s refers to the translation parameter of the wavelet, and t refers to time.

4. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 1, characterized in that: In the registration module, the deep effect brain region is defined by: The inverse transformation matrix of the MNI152 standard space template was used to transform the preset coordinate spherical ROI into the individual space.

5. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 1, characterized in that: In the calculation module, the functional connectivity calculation adopts the following Granger causality model: Where: X t and Y t are the wavelet coefficients of the deep effect brain area and cortical voxels respectively; p represents the hysteresis coefficient; A k B is the historical influence weight of deep brain area X on cortex Y; k is the autoregressive weight of the historical state of cortex Y itself; A ' k is the historical influence weight of cortex Y on deep brain area X, B ' k is the autoregressive weight of the historical state of deep brain region X itself; C and C ' are the regression coefficients of the covariates; E t and E t ' are error terms; Z t is a covariate.

6. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 1, characterized in that: In the target localization module, the top clusters were screened at an interval of 5%, and the voxels with the top 10%-50% Granger causality values ​​in the left dorsolateral prefrontal cortex were selected. The cluster boundary value was set to 26 and the number of voxels was greater than 20. The center of gravity of the cluster under each threshold was calculated as the candidate target.

7. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 6, characterized in that: The centroid is calculated as the weighted geometric centroid: Where: X i is the 3D coordinate of the voxel, n j is the total number of voxels contained in the jth cluster; w i Each voxel coordinate value is weighted by the Granger causality value of its voxel.

8. The transcranial magnetic stimulation positioning system based on deep effect brain areas according to claim 7, characterized in that: The final stimulation target must be located at the center of gravity of the top 20% functional connection cluster in the Ranger causality value and be less than 30 mm away from the skull surface.

9. The transcranial magnetic stimulation positioning system based on deep effect brain regions according to claim 1, characterized in that: The filtering unit adopts a multi-scale credibility optimization strategy, specifically including: Frequency band selection unit: preferentially selects the 0.06-0.125 Hz frequency band for functional connectivity calculation; Basis function verification unit: dynamically selects the optimal basis function by comparing the reliability index ICC ≥ 0.4 voxel number of Daubechies db2 / db4 / db6 basis functions.

Images