5) Analysis of Anatomical Location of Cluster Endpoints

Commands to do endpoint analysis:

The endpoint analysis allows users to find fiber clusters according to their endpoints. This needs the users to provide information about the brain parcellation, such as freesurfer cortical parcellation and fMRI functional areas, as format of label map. The method will compare the cluster's endpoints to each region in the label map and then identify the clusters that are robustly connected to a region-of-interest across multiple subjects.

wm_harden_transform.py

• This script will apply a transform in itk format to a vtk polydata (tractography) file.
• Use this command to transform fiber clusters from the atlas space to DWI space (individual subject space), to then be able to compare the fiber tract endpoints with subject-specific brain regions. 3D Slicer is used to perform the transform.
• This requires that the label map defining the brain regions has been registered to the DWI data. For example if using Freesurfer, the parcellation and its corresponding T1/T2 must be aligned to the DWI data.
• To transform fiber clusters to subject space, multiple transforms may be needed if both affine and nonrigid registrations were conducted when generating the atlas. (If this is the case, be sure to work backward by first applying the inverse of the nonrigid transform followed by the inverse of the affine transform.)
• -i needs to be specified for inverse transform to invert the transform files (XX.tfm) generated from the atlas registration. (Note these transforms registered the individual subject data into atlas space, so to reverse the process the inverse transforms are needed.)

wm_harden_transform.py subject_cluster_folder output_transformed_cluster_folder path_to_slicer -t registered_tractography/output_tractography/itk_txform_SUBID.tfm -i

wm_measure_endpoint_overlap.py

• Computes overlap of each fiber cluster's endpoints to the regions in the label map. This will give a percentage per cluster per region, which shows how many endpoints of the cluster touch the region.
• This can measure multiple subjects at the same time.

wm_measure_endpoint_overlap.py all_subject_fiber_clusters_folder all_subject_label_map_folder output_measurement_result_folder path_to_FiberEndPointFromLabelMap

wm_extract_clusters_by_endpoints.py

• Extract fiber clusters that connect to a region-of-interest.
• First, you need to define a percentage threshold (-p) to decide if a cluster connects the region-of-interest in individual subject. For example, cluster C has 10% endpoints connect to the region-of-interest R; if the percentage threshold is 20%, C and R will be considered as disconnected.
• Then, you need to define a subject number threshold (-s) to decide if a cluster connects the region-of-interest robustly across multiple subjects. For example, for a dataset of 50 subjects, cluster C connects to the region R in 45 subjects under the percentage threshold 20%; if the subject number threshold is 48, C and R will be considered as disconnected in the population.
• The output will be the clusters that pass the two thresholds.
• The script will generate mrml files to view the extracted clusters for each subject, if -fc_folder is specified.

wm_extract_clusters_by_endpoints.py output_measurement_result_folder -p PERCENTAGE_THRESHOLD -s SUBJECT_NUMBER_THRESHOLD -fc_folder FIBER_CLUSTER_FOLDER