CN-116385376-B - Brain region target point dividing method for autism nerve regulation and control

CN116385376BCN 116385376 BCN116385376 BCN 116385376BCN-116385376-B

Abstract

The invention discloses a brain region target division method for autism nerve regulation, which comprises the steps of obtaining resting state functional magnetic resonance imaging data of an autism patient, preprocessing the functional magnetic resonance imaging data, extracting dorsal lateral forehead cortex ba9 and ba46 in the preprocessed data by using a Brodmann template as target brain regions, expanding the preprocessed image data into a two-dimensional time sequence, carrying out sliding window processing on the time sequence to obtain a plurality of sliding window time sequences, calculating a functional connection matrix of voxel points of the target brain region in each sliding window time sequence and all voxel points in whole brain grey matter, carrying out dimension reduction processing on all the functional connection matrices based on local similarity to obtain a dimension reduction functional connection matrix, obtaining a constructed reference clustering template of normal personnel, and carrying out Kmeans clustering on the functional connection matrix of the autism patient according to the clustering cluster number of the functional connection matrix to obtain the subarea division of the target brain region.

Inventors

LI YANLING
GAO JINGJING
LI RUI

Assignees

西华大学
电子科技大学

Dates

Publication Date: 20260512
Application Date: 20230317

Claims (7)

1. The brain region target division method for autism nerve regulation is characterized by comprising the following steps: S1, acquiring resting state functional magnetic resonance imaging data of an autism patient, and preprocessing the functional magnetic resonance imaging data; S2, extracting dorsolateral forehead cortex ba9 and ba46 in the preprocessed data by using a Brodmann template to serve as a target brain region; s3, spreading the preprocessed image data into a two-dimensional time sequence, and performing sliding window processing on the time sequence to obtain a plurality of sliding window time sequences; s4, calculating a functional connection matrix of voxel points of a target brain region in each sliding window time sequence and all voxel points in the whole brain gray matter; S5, performing dimension reduction processing on all the functional connection matrixes based on the local similarity to obtain a dimension reduction functional connection matrix; s6, acquiring a constructed reference clustering template of normal personnel, and carrying out Kmeans clustering on the functional connection matrix of the autism patient according to the clustering cluster number to obtain subarea division of a target brain region; The method for constructing the reference clustering template of the normal personnel comprises the following steps: a1, acquiring resting state functional magnetic resonance imaging data of normal personnel from different sites, and preprocessing; a2, acquiring a dimension reduction function connection matrix of each normal person in a mode of S1-S5; a3, fusing the dimension reduction function connection matrixes of all normal people based on weighted multi-source data fusion to obtain a final function connection matrix; a4, calculating Euclidean distance between voxel points in the final functional connection matrix as a similarity matrix, and performing group level AP clustering on the similarity matrix to obtain a group level initial cluster; and A5, finding the maximum clustering contour of the group horizontal initial clustering according to the contour coefficient and Calinski Harabaz Index, and obtaining the brain region of the normal personnel group horizontal subarea division as a reference clustering template.
2. The method of brain region target division for autism neuromodulation according to claim 1, wherein step S5 further comprises: s51, calculating Euclidean distances between each row in each functional connection matrix, and arranging all Euclidean distances into a square matrix to form a functional connection similarity matrix; s52, combining all the function connection similarity matrixes of the same normal person/autism patient to obtain a dynamic function connection similarity matrix; s53, subtracting two adjacent layers in the dynamic function connection similarity matrix to obtain an interlayer difference, defining a mask matrix, and traversing each layer of the interlayer difference by using the mask matrix in a preset step length; s54, calculating all positive values in each mask area to form a positive value gradient matrix, and calculating all negative values to form a negative value gradient matrix; S55, extracting the maximum value and the minimum value in all positive gradient matrixes and negative gradient matrixes of the same interlayer difference, squaring, and then respectively finishing the squares into a local gradient increasing matrix And a local gradient reduction matrix ; S56, calculating variances of all positive gradient matrixes and variances of all negative gradient matrixes of the same interlayer difference to respectively form a positive variance matrix and a negative variance matrix, and normalizing to obtain a normalized positive variance matrix Sum-of-normalized negative variance matrix ; S57, adding matrix to local gradients respectively And a local gradient reduction matrix Performing first-order linear fitting on the window time sequence of each element in the sequence to obtain a dynamic change trend matrix And ; S58, respectively to 、、 And Reconstructing to obtain reconstructed matrixes respectively 、、 And ; S59, each interlayer difference is corresponding to the interlayer difference 、、 And And carrying out convolution to obtain a dimension reduction matrix, and accumulating all dimension reduction matrices to obtain a dimension reduction functional connection matrix.
3. The method for dividing brain region targets under the regulation of autism nerves according to claim 2, wherein the method comprises the steps of, And The calculation formulas of (1) are the same, wherein The calculation formula of (2) is as follows: Wherein, the The dynamic change trend matrix of the e-th interlayer difference; is the g value in the e-th layer gradient matrix; is a layer e time series; For all layers The average value; for all time-series layers The average value; The total number of the sliding window time sequences; The calculation formula of the dimension reduction function connection matrix is as follows: Wherein, the Is the e-th interlayer difference; 、、 And Respectively corresponding to the e-th interlayer difference 、、 And And reconstructing the matrix.
4. The method for brain region target division of autism neuromodulation according to claim 2, wherein the positive gradient matrix and the negative gradient matrix, wherein the calculation formula of the positive gradient matrix is: Wherein, the A positive gradient matrix for the o mask in the e-th interlayer difference; And The last value and the previous value of the x-axis in the o-th mask are respectively the difference between the e-th layers; And The last value and the previous value of the y-axis in the ith mask are respectively the difference between the ith layer and the ith layer; An X-direction gradient component of an o-th mask in the e-th interlayer difference; is the Y-direction gradient component of the o-th mask in the e-th interlayer difference.
5. The brain region target division method for autism neuromodulation according to claim 1 or 2, wherein the method for fusing the dimension reduction function connection matrix of all normal persons based on weighted multi-source data fusion comprises the following steps: A31, merging the dimension reduction function connection matrixes of all normal persons to form a [ m, m and D ] three-dimensional function connection matrix Fm, wherein m is the number of voxels in a target brain region, and D is the total number of normal persons; A32, removing abnormal values in the three-dimensional function connection matrix Fm, and then carrying out standardization processing on each voxel of each normal person in the Fm matrix: Wherein, the The function connection of the (r) th voxel and the (t) th voxel of the (a) th normal person after standardization; The r-th voxel of the a-th normal person is functionally connected with the t-th voxel; Functional connection of the nth voxel to the nth voxel for all normal persons involved in the study; a33, dividing the Fm matrix into m matrices with the size of [ m, D ] along the X axis of the Fm matrix Calculating Specific gravity of functional connections of all normal persons for each site in the matrix : , Wherein, the Is the (u) th voxel The average functional connection of each site accounts for the specific gravity of the functional connection of all normal personnel, and the size is S, m; is the first The functional connection value of the first normal person of the site on the X axis is L which is the total number of normal persons of the S j site; Is that Middle (f) All voxels of each site have an average functional connection matrix of size [1, m ]; Representing a matrix of size S, m, wherein ; A34, calculating the average function connection of each site Uncertainty of (2) : Wherein, the The size of (2) is [1, S ]; a35, according to the functional connection uncertainty of each site Calculating the functional connection weight value of the current voxel : Wherein, the The size of (2) is [1, S ]; a36, connecting weights according to all functions of each site Calculation of Voxel function connection matrix composed of all normal persons of matrix : Wherein, the A functional connection matrix which is the t-th voxel of the functional connection matrix X axis; The size of (2) is [1, m ]; A37, dividing the Fm matrix into m matrices with the size of [ m, D ] along Y axis of the Fm matrix Calculating a matrix by adopting the mode of the steps A33-A36 All voxels of the matrix ; A38, all of the X-axis Combining to form a matrix For all of the Y-axis Combining to form a matrix ; A39 is formed into matrix according to pairing And Calculating final function connection matrix : Wherein, the Is that Is a transpose of (a).
6. The method for brain region target division under autism neuromodulation according to claim 2, wherein the method for preprocessing functional magnetic resonance imaging data comprises: B1, removing N time point images before the functional magnetic resonance imaging data of the autism patient/each normal person; b2, registering the functional magnetic resonance imaging processed in the step B1, and removing data with the head movement of more than 2mm or 2 degrees in any direction; b3, matching the image processed in the step B2 to an EPI template by adopting the EPI template, resampling to the voxel size of 3 x 3, and removing the linear trend; B4, carrying out band-pass filtering on the data with the linear trend removed by 0.01-0.1 Hz, and then adopting regression Friston's-24 head movements, white matter, cerebrospinal fluid and whole brain average signals in DPARSFA software; and B5, removing low-quality data with the average frame displacement larger than 0.5 from the data processed in the step B4, and performing smoothing processing by adopting an 8mm full-width half maximum Gaussian check image.
7. The brain region target division method for autism neuromodulation according to any one of claims 1-4 and 6, wherein the calculation formula of the functional connection matrix in step S4 is: Wherein, the All voxel point time sequences of the ith sliding window time sequence in the target brain region; Is a time sequence of all voxel points in the gray matter of the whole brain; The standard deviation of all voxel point time sequences of the ith sliding window time sequence is obtained; the standard deviation of the time sequence of all voxel points of the whole brain gray matter; (.) is covariance function, m is the number of target voxel points, and n is the number of gray matter voxel points of whole brain.

Description

Brain region target point dividing method for autism nerve regulation and control Technical Field The invention relates to a brain region dividing method, in particular to a brain region target dividing method for autism nerve regulation and control. Background The brain is the finest system in the human body, a large number of neurons participate in the formation of a cerebral neural network, and autism is a typical abnormal disease of the cerebral neural network, and the pathogenesis of the autism often involves abnormal functional (neural) connection among a plurality of brain regions. When the autism nerve regulation and control is studied, researchers mainly concentrate effective treatment targets on DLPFC (dorsal lateral prefrontal cortex) on two sides of the brain, but the cortex area covered by DLPFC on two sides of the brain is too large, and different individuals are inconsistent in brain development condition, and focuses of the individuals are not uniform, so that the current researchers cannot accurately position the stimulation targets of different study objects when the autism nerve regulation and control is studied, and the targeting reference of scientific researchers is inaccurate. Disclosure of Invention Aiming at the defects in the prior art, the brain region target point dividing method for autism nerve regulation and control solves the problem that the brain region of an autism patient cannot be accurately divided in the prior art. In order to achieve the aim of the invention, the invention adopts the following technical scheme: the brain region target division method for autism nerve regulation is provided, and comprises the following steps: S1, acquiring resting state functional magnetic resonance imaging data of an autism patient, and preprocessing the functional magnetic resonance imaging data; S2, extracting dorsolateral forehead cortex ba9 and ba46 in the preprocessed data by using a Brodmann template to serve as a target brain region; s3, spreading the preprocessed image data into a two-dimensional time sequence, and performing sliding window processing on the time sequence to obtain a plurality of sliding window time sequences; s4, calculating a functional connection matrix of voxel points of a target brain region in each sliding window time sequence and all voxel points in the whole brain gray matter; S5, performing dimension reduction processing on all the functional connection matrixes based on the local similarity to obtain a dimension reduction functional connection matrix; s6, acquiring a constructed reference clustering template of normal personnel, and carrying out Kmeans clustering on the functional connection matrix of the autism patient according to the clustering cluster number to obtain the subarea division of the target brain region. The method has the beneficial effects that the method combines the constructed reference clustering templates, finely divides the brain region based on the functional connection matrix of the voxel points, finds out the subregion with obvious difference of the brain neural network of the autism patient, and is used as a key research area of scientific researchers, so that errors caused by manually positioning the stimulation target can be effectively avoided, and the accuracy of research on the disease cause of the autism patient is ensured. Further, the method for constructing the reference clustering template of the normal personnel comprises the following steps: a1, acquiring resting state functional magnetic resonance imaging data of normal personnel from different sites, and preprocessing; a2, acquiring a dimension reduction function connection matrix of each normal person in a mode of S1-S5; a3, fusing the dimension reduction function connection matrixes of all normal people based on weighted multi-source data fusion to obtain a final function connection matrix; a4, calculating Euclidean distance between voxel points in the final functional connection matrix as a similarity matrix, and performing group level AP clustering on the similarity matrix to obtain a group level initial cluster; And A5, finding the maximum clustering contour of the group horizontal initial clustering according to the contour coefficient and CalinskiHarabaz Index, and obtaining the brain region of the normal personnel group horizontal subarea division as a reference clustering template. The technical scheme has the beneficial effects that the result obtained based on the large sample data of the multiple sites is more reliable than the result obtained based on the small sample, the time characteristic of the data is considered, the calculation complexity is reduced, and the clustering result based on the voxels is finer. Further, step S5 further includes: s51, calculating Euclidean distances between each row in each functional connection matrix, and arranging all Euclidean distances into a square matrix to form a functional connection similarity matrix; s