Hi,
After processing the exact same subject with the exact same pipeline on two different machines and finding totally different results i discovered that the incriminating step was the SPM tissue segmentation which seems to dramatically influence the final output depending on Matlab version (R2009A [failure] vs R2013A [somewhat OK]).
Could someone please comment on this? This is a serious issue if SPM output is highly dependent on Matlab version.
Here is a link to the test case data https://www.dropbox.com/sh/tp9nxmajf2x8efm/hQeg6hyLEY/SPMsegMATLABver with:
- acos007a1001.{hdr,img} : the subject T1
- {white,grey,csf}_1mms.nii : the SPM standard space tissue probability maps resliced to 1mm
- data.rar : all the previous files compressed
- seg_spm8_testcase.m : the script taking as input the path to the directory containing the previous files and computing the segmentation (as in the pipeline i analyzed)
- R2009Aoutput.rar : the segmentation output from SPM in Matlab R2009A
- R2013Aoutput.rar : the segmentation output from SPM in Matlab R2013A
Thanks,
Michael
|