Thank you for your detailed reply, John. I tried SPM12, but I'm afraid the segmentation is still not satisfactory. Probably the raw space T1 image is far from well aligned with MNI space.

The attached images are a quick demonstration of what I was referring to, all performed in SPM12.

The raw.png is my raw space T1 image, and the raw-gmpm.png is the native space GM probability map given by 'Spatial'->'Segment' module in batch editor.

The coreg.png is from the same subject after linearly aligned to MNI space, and the coreg-gmpm.png is the GM probability map of the aligned image, which is much more reasonable.

Could you please suggest what may cause this and advise how I can run DARTEL on the raw space image and avoid running it on the aligned image? Thanks again.

Shengwei