Take a look at the original scans and you'll see that the extremities of the brain are not included in the FOV. This is a result of those parts of the original images not being present. Instead of assuming that the values are zero in those regions (which is wrong), it takes the values of the closest piece of available data (which is also wrong). When you run the stats, you just need to ensure that this part of the data is masked out. Best regards, -John On 21 July 2011 15:43, Peng Syu-Jyun <[log in to unmask]> wrote: > Dear Sir > > After DARTEL process to normalise to MNI space, but why > produce the unwanted parts in the red circuit? How could I avoid? > > Best regards, > > Syu-Jyun