Hackpads are smart collaborative documents. .

Tor Wager

923 days ago
Unfiled. Edited by Tor Wager 923 days ago
Tor W *The image above is flipped left/right*
SPM uses neurological convention: the Left is on the Left. Lots of data are collected in radiological convention (the Left is on the Right). SPM isn't going to tell you if it is flipping or not, however the X-coordinate will be negative if SPM thinks it is in radiological and is is therefore flipping it for your viewing convenience. You should mark with vitamin E the Right side of all subjects. If you haven't done this, in order to confirm that what you are analyzing as R side in SPM actually is R side in real life, consider natural asymmetries in the brain: The right hemisphere protrudes anteriorly beyond the left, and the left hemisphere extends posteriorly beyond the right (Toga & Thompson, 2003).
Preprocessing QC checklist
  • Images: Confirm that mean imgs from each run look like brains
  • Dropout: Look at SNR across the brain.  Expect it to be lower in certain areas (i.e. OFC), but know that areas with SNR < about 30 could be quite problematic
  • Orientation: Check orientation -- confirm L is L (occiptal will be larger on L)
  • Raw data quality: Look at movie of raw data:  Successive differences movie (this is not suppressed in luka's wrapper) :: way of identifying gradient artifacts in raw data. 
  • Head motion: Check that head motion is minimal (one rule of thumb is < 2mm overall or < 1mm within run)
  • Coregistration: For each subj, check structural-functional coreg visually (our preproc code makes you do this, too)
  • More image registration: confirm that warped anatomical, warped mean func, and template are all in register (normalization.png?)
  • Normalization: scnlab_norm_check to check normalization (takes template (avg152T1.nii), should be copied to a directory where you have permissions, warped T1s for each person, mean functional files, and subject folders as cell array)
  • Individual subject brain segmentation: check that each subject's mask looks like expected

Contact Support

Please check out our How-to Guide and FAQ first to see if your question is already answered! :)

If you have a feature request, please add it to this pad. Thanks!

Log in