Tissue Segmentation and Tissue Priors
Following generation of the template, the ANTs cortical thickness pipeline was run to accomplish tissue segmentation on the template. In order to run tissue segmentation on the previously generated template, a standard template must first be obtained. A template consisting of male and female participants, ages 10 to 18, was downloaded online (
https://s3-eu-west-1.amazonaws.com). After completion of the download, the file was unzipped and the original zip file removed.
Before analyzing individual participants, tissue segmentation must first be accomplished on the template image through the pipeline. In addition, several files necessary for individual tissue segmentation were generated from the program. A whole brain ROI mask was created, isolating brain matter in the field of view. This mask was used in generating an extracted brain image, eliminating skull, dura mater, and any other non-brain data. The whole brain ROI mask was next converted to a probability mask. which will be overlain across brains to determine the probability of each voxel corresponding to cerebrospinal fluid, white matter, or gray matter. The ROI mask image was then dilated to 28 voxels isotropic to create an extraction mask, and 18 voxels isotropic for a registration mask.
Through the ANTs cortical thickness pipeline, brain tissue was segmented into 6 specific regions of interest (ROIs). Intensity values of the ROI image ranged from 1 through 6, and corresponded in ascending order, respectively, to cerebrospinal fluid, gray matter, white matter, the basal ganglia, the brain stem, and the cerebellum. Six tissue priors were created, corresponding to each of the six ROIs converted into probability masks. Accomplishing this tissue segmentation through ANTs was essential in order to highlight the cortex and strip away structures that were not of interest.