Hello, We recently just ran 80 elderly brains through -autorecon1 using FS v4.0.2, and encountered 3 different types of errors that we are not entirely sure how to handle:
(1) During mri_normalize of 1 subject we encountered: SPLINE_CUBIC_SET - Fatal error! The number of data points N must be at least 2. The input value is 1.
The FS archive indicates to check the talairach transform given this error, but the talairach.xfm appears fine for this subject (as assessed using tkregister2).
(2) During mri_watershed of 1 subject: Sorting... mri_watershed Error: Problem in the COG calculation
This subject had a slightly warped talaraich.xfm, but I think more relevantly, its talairach_with_skull.lta was way off.
(3) During mri_watershed of 5 subjects: mri_watershed Error: GLOBAL region of the brain empty !
For all 5 of these subjects, the talairch_with_skull.lta also had obvious problems.
RE (1): Any insight would be appreciated.
RE (2) and (3): I suspect that we should try running with the -no- wsgcaatlas flag, so that talairach_with_skull.lta is longer used by mri_watershed. Should we then substitute the -wsatlas flag instead? (Its not clear to me which atlas is used when that particular flag is included).
Last, I thought you might be interested in these errors, because they directly relate to my recent question regarding the robustness of talairach_with_skull.lta. At least during the -autorecon1 (i.e., initial) creation of this lta file, it appears that there is some possibility of a bad talairach_with_skull registration (6 of 80 in this particular subject set).
thanks, Mike H.