Dear Mrtrix experts,
I would like to perform connectivity based parcellation applying k-means clustering algorithm. I already carried out connectivity based parcellation with a hypothesis driven approach employing whole brain tractography.
I am looking for employing k-means to do a hypothesis driven/data driven comparison. If I well understood, I need to obtain a matrix computing tractography between each voxel of my seed region and all target regions, then apply k-means algorithm to obtain a number k of non-overlapping clusters derived from similarity of connectivity profiles among voxels.
I read previous discussions on this forum and I would ask you if this pipeline is suitable to achieve what I am looking for:
- Use maskdump with the seed region to obtain voxel location
- Use mredit to obtain single voxels ROIs
- Use tckedit filter out connectivity profiles of each voxel from the whole brain tractography
How can I obtain the matrices requested to perform k-means clustering (i.e. on MATLAB)?
Is this correct? are there any ways to obtain matrices more properly?
Thank you in advance for any help or precious suggestion