 |
 |
Hi, It is likely that your data is either noisy or contains white matter hypointensities (often occur due to age or disease) that are being classified as non-white-matter. So it could be that fast 4 is more or less sensitive to noise or "signal". I can't tell from the images I've seen. Either way, what I said before is true in that the default parameters are OK for the kind of average scans we've tested on but it is always worth looking at the results carefully for your own sequences and subjects and then adjusting the parameters appropriately. All the best, Mark On 4 Feb 2011, at 10:41, Andreas Glatz wrote: > Mark, > > Thanks a lot for the tip. > > I played around a bit and found that '-f 1.08' and '-H 0.3' give much better results with fast v417 and my data. > > Of course, now I'd like to know why I have to change the default values of those parameters which are: > > -f initial segmentatioin spatial smoothness; default=.2 > -H segmentation spatial smoothness; default=.1 > > Is it because my data is rather noisy and fast has to smoothen it a lot to get a good estimate of the bias field? > > And there is still that other question: > > Why didn't I ran into a similar problem with fast v353? > > I am happy for any hint to answer those questions. > > Best regards, > Andreas > >