 |
 |
Dear Mishkin, If it is bright white matter that is the problem then it may be related to something we've seen before - if the CSF is quite noisy then the CSF Gaussian model tends to have a very large variance, making it extend into the high intensity range with a non-negligible probability. The consequence is that it can cross over with the white matter Gaussian at the extremes of the white matter intensity range. This is obviously not desirable, but is a consequence of the simple mathematical model which is at the heart of FAST. One solution we've got to work in the past is to increase the number of classes to 4, which can result in the CSF getting modelled in two classes, as often there are some very dark portions of the CSF and some less-dark, and these are more effectively modelled by two Gaussians, limiting the extent of their reach into the high intensities. So try this and let us know how you get on. As a last resort you can always take the white matter mean intensity and threshold the "CSF" component to exclude the brighter voxels, but try the 4-class solution first. All the best, Mark On 28 Feb 2011, at 19:51, Mishkin Derakhshan wrote: > I just noticed my critical typo. They are hyperintensities (high). > > The images do have some white matter hypointensities (MS lesions), but > I've masked them out (in reality, sienax -lm has masked them out for > me). > > > On Mon, Feb 28, 2011 at 11:38 AM, Mark Jenkinson <[log in to unmask]> wrote: >> Dear Mishkin, >> Could you let me know whether the areas what are >> excluded from the white matter correspond to high >> intensities (hyper) or low intensities (hypo)? I struggled >> to see what it was from your images and description, >> and they would represent really very different problems. >> All the best, >> Mark >> >> >> On 25 Feb 2011, at 19:32, Mishkin Derakhshan wrote: >> >> Hi, >> Like a few others on the list I've had problems with FAST v4.1 leaving >> holes in the segmentation ie. classifying parenchyma as CSF. >> >> http://i.imgur.com/7KLqh.png >> >> Looking at the data (see above link for examples), it seems like this >> might be related to white matter hypointensities that are not disease >> related, and is present at the mixeltype stage. >> >> I've tried varying f, H, R, capital i and lower case L and N, but I've >> had little success. >> My best attempts have been lowering the bias field smoothing kernel to >> a FWHM of 4, but this still leaves/creates many small holes, and it >> drastically changes the results of the GM segmenation. >> >> Any tips on what flags I should vary and in what direction (ie. >> bigger/smaller than the default) to help this? >> >> thanks, >> mishkin >> >> >> >