Hi AFNI gurus,
I don't quite understand what 3dkmeans does by simply looking at the help. With an input set of "-f F1+orig F2+orig F3+orig F4+orig", does it classify input files so that F1/F2 are grouped together and F3/F4 are grouped in a different category; or, does it classify voxels in each input files and produce 4 different classification results; or, does it consider F1 through F4 as the temporal features of a single volume and performs classification based on these temporal features thus produce only one parcellation result? What is exactely the "feature" that this kmeans classifier works on?