I am doing the first level ICA analysis using Melodic on my T1w and resting-state fMRI data from patients with DBS. Data of most subjects underwent the process without problem, but a few of them failed the registration between 3D T1 and standard MNI152. The registered brains were distorted and rotated in a wrong way. I have tried the options of Nonlinear and FullSearch but it didn’t help. I have checked both T1w and fMRI image before process and both of them looks fine.
Try running fslreorient2std on all of your data before doing anything else. Doing this will substantially reduce the search space that FLIRT has to navigate to find a good alignment between the two images.
The fslreorient2std does correct the orientation problem see image below. However, the issue of the extreme distortion remains as before. I wonder what can cause the excessive distortion during registration between T1 and MNI152. Is there anything I can do to make it better?
The distortion issue have been resolved when performed fslreorient2std on the normalized T1 image from Freesurfer and use it as anatomical image in Melodic.
