I think I know why it is crashing. The function that works on the connected
voxels of each blob is recursive and keeps pushing and popping data from
the stack. If the blobs are very large, the stack grows to beyond its limit,
and causes Matlab to crash. I would guess that for your particular contrast,
you have some very large blobs - possibly indicating global volume differences.
The way around the problem would be to increase the stacksize. If you are
running Matlab from a Unix machine, you should type the following before
invoking it:
unlimit stacksize
Best regards,
-John
| We are using the determinant of the Jacobian of the normalization warps
| of
| structural MRIs in a multi-subject analysis with age as the covariate
| of interest (three subjects with 84 images each; please see attached
| bitmap
| of the spm design matrix).
|
| On defining 18 t-contrasts, two of these (-1 0 0 0 0 0 and 0 0 -1 0 0
| 0)
| result in a segmentation fault causing the matlab session to core dump.
| No
| error message appears either in spm or in matlab. The corresponding
| images
| con_000x.img and spmT_000x.img are however written out before the core
| dump
| occurs. Also these images can be viewed without trouble using the
| spm_display tool.
| A similar analysis of smooth normalized gray images does not result in a
| core dump.
|
| Any pointers as to the cause of this spm/matlab crash would be greatly
| appreciated.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|