Hello,
I was wondering if anyone could help me. I am using an Open access dataset downloaded from OpenNeuro that provided fmriPrep multi-band minimally processed resting-state fMRI data. My aim is to run a Group-ICA in GIFT using this data. The preprocessing steps followed are:
- fMRIprep (distortion correction, brain extraction, registration, slice timing correction, normalisation…)
- Removal of first two volumes
- Spatial smoothing
I then proceeded to run a Group ICA, however most of my components have these large jumps at the start of the time courses (see image) along with large PowerLF/PowerHF values.
I would be less concerned if this was only a few components however the majority of my components have this spike. I had originally included high pass filtering (0.01Hz) in my processing protocol but when I ran the ICA with this, MDL optimal component estimation gave 1 component and when I forced a 20 model order decomposition all of the resulting components were noise (ventricles or susceptibility areas).
Does anyone have any advice on what might be going wrong here? Am I missing a core processing step or is my data incompatible with ICA?
Thank you in advance for any help.