Dear Masahiro, > I am working on MRIs with lesions, which may be misclassified as gray > matter in the process of segmentation. I need gray matter images to do > partial volume correction of PET data. > > By searching the SPM mailbase, I found the communications copied at the > bottom. However, http://www.medicalimagecomputing.com/EMS/, shown in the > communication back in 2004 cannot be accessed anymore. Could you provide > a link if it is still useful with SPM5? A quick google search gives the new address as: http://www.medicalimagecomputing.com/downloads/ems.php?EMSsection=about If I remember correctly, it requires SPM99 - which may be a bit tricky to get working with more recent MATLAB versions. > > Is there a better method available now? A number of more recent lesion segmentation algorithms have been developed, but I am not aware of any implementations that have been made freely available. The algorithms are published in the literature, but generally there are no incentives for anyone with the right skills to bother re-implementing and supporting them. This will be the case for as long as the quantity of journal publications remain the main measure of scientific productivity. > > In addition to T1-weighted images, I have FLAIR images. The intensity of > the lesions in T1 is somewhat like that of gray matter. However, the > intensity is quite high in FLAIR. Multi echo segmentation may just be > sufficient for my purpose of partial volume correction. I know multi > echo segmentation is operational in SPM2. Is it operational in SPM5 now? The old multi-channel segmentation code of SPM2 is still there in SPM5, and should be available via the Toolbox pulldown. I am currently working on getting multi-channel segmentation up and running for the next SPM release - although it may require quite a bit of customisation to get it to accurately segment lesions. > I am asking this question because sometime ago, I saw a communication > saying that multi echo segmentation was not operational in SPM5. There is no SPM5 implementation of "Unified Segmentation" that takes multiple channels - but the older segmentation algorithm is still available. Best regards, -John