- Theory
The structural covariance network (SCN) captures shared morphological covariance patterns between brain regions according to morphological measures derived from structural magnetic resonance imaging (sMRI). Traditional group-level SCN approaches consider subjects as time points, which are largely influenced by the participant size and hard to characterize interindividual heterogeneity. Thus, several individualized SCN (ISCN) were proposed to sketch the individual covariance variability, such as morphometric similarity networks (MSN), Kullback-Leibler divergence-based morphological brain network (KLD-MBN), and individual differential structural covariance network (IDSCN). We developed a novel individualized similarity network (TSN) based on hundreds of sMRI-derived 3D voxel-wise texture feature maps (for example, GLCM), which has refined spatial information and has been proved to provide additional covarying information relative to other ISCNs. The pipeline for TSN is shown in the following:
- Usage
This package processing as follows:
Step 1: All sMRI data (T1-weighted brain image) were preprocessed using the CAT12 toolbox (http://dbm.neuro.uni-jena.de/cat/) implemented in SPM12 (http://www.fil.ion.ucl.ac.uk/spm/). And then, the native gray matter (e.g., p1*.nii) and white matter (e.g., p2*.nii) concentration maps were obtained for each subject using the CAT12 toolbox. In addtion, the inverted nonlinear deformation-field parameter (e.g., iy_*.nii) tranforming from the standard space to the individual native spce was remained for the following the individual-level network construction.
Step 2: The standard whole brain atlas (e.g., BNA_maxprob_thr25_1.5mm.nii, 246 cerebral parcellations from Human Brainnetome Atlas) was pre-defined as the nodes of TSN for the following the individual-level network construction.
Step 3: Voxel-wise GLCM texture maps were carried out using a self-developed program (e.g., batch_vox_glcm.sh) based on Matlab 2016b. A total of 180 texture feature maps was ultimately generated from the 9 typies of sMRI data (1 original and 8 wavelet-transformed brain maps) for each subject. This script contains 6 inputs as follows: (1) T1 rawdata map (corrected for bias-field); (2) the native gray matter map; (3) the native white matter map; (4) the result directory for saving texture feature maps; (5) Subject ID; (6) the full path of the package.
E.g., batch_vox_glcm.sh ./Subj001/mT1.nii ./Subj001/p1T1.nii ./Subj001/p2T1.nii ./result/Subj001 Subj001 ./TSN-VGLCM
Step 4: TSN construction were also carried out using a self-developed program (e.g., batch_vox_glcm_matrix.sh) based on Matlab 2016b in Linux system. This script constructed the subject-level brain TSN using the pre-defined standard whole brain atlas from Step 2. The standard brain atlas was first warped into each subject's native space using the nonlinear deformation warp map (e.g., iy_*.nii) generated at Step 1. Then the feature vector of each parcellation of each subject was extracted from the 180 VGLCM maps. After that, a Pearson correlation was used to calculate the covariance coefficient of the feature vectors between each pair of areas. Then Fisher's r-to-z transformation algorithm was used to convert the covariance coefficient to approximately normally distributed, resulting in a symmetrical covariance matrix (termed TSN). This script contains 6 inputs as follows: (1) the directory for saving texture feature maps from Step 3; (2) the directory for saving T1 rawdata map and the nonlinear deformation warp map; (3) the full-path filename of the standard brain atlas; (4) the result directory for saving TSN; (5) Subject ID; (6) the full path of the package.
E.g., batch_vox_glcm_matrix.sh ./result/Subj001 ./Subj001 ./TSN-VGLCM/BNA_maxprob_thr25_1.5mm.nii Subj001 ./TSN-VGLCM
Reference: [1] Ding H, Zhang Y, Xie Y, Du X, Ji Y, Lin L, Chang Z, Zhang B, Liang M, Yu C, Qin W, Individualized texture similarity network in schizophrenia, Biol Psychiatry, 2024, doi:10.1016/j.biopsych.2023.12.025