I've got strange results with melodic on brain-extracted, pre-processed BOLD data. I'm attaching the first component returned for both raw (no brain extraction) and brain-extracted data. The time course of the first component in the raw data exactly matches the paradigm we used. However, after applying the brain mask (using an in house matlab code) the time course seems to be inverted and red and blue regions are swapped (compared to non-brain extracted case). I'd appreciate any thoughts on what might have gone wrong.