Dear Irina,
A distorted beta image is rather difficult to explain as in principle, a single volume with a certain artefact might be sufficient to affect the beta estimate if it coincedes with the predicted time course of that regressor.
I'd suggest to look at the raw data. Some scanner / coil artefacts might show up in particular slices only, if you realign & normalise data this might not be obvious any more. E.g. sometimes, the first few volumes will show stripes with some slices having a higher intensity than others. ArtRepair http://cibsr.stanford.edu/tools/human-brain-project/artrepair-software.html might be a good try, as it has a movie function displaying all the slices of a volume at the same time, plus there are options to increase and color-code the contrast.
Best
Helmut
|