> I would like to normalize T2 spin echo images of patients with brain
> lesions. I don't have any T1-weighted images for these patients. I was
> wondering if the unified segmentation in SPM5 could work with T2-weighted
In principle, it segmentation should work with T2w scans. Lesions are not
handled so well though, and the masking is the best that can currently be
> I tried on some patients, and visually it was not bad, but segmentation of
> grey/white matter from T2 images may be controversial...
The current "gold standard" with which automatic algorithms are compared is
manual segmentation. Therefore, the best way of judging the accuracy of the
segmentation is probably by visual examination.
> Does somebody ever tried it, or has any advice about which method to use
> (in particular lesion masking in SPM2 of unified segmentation in SPM5)?
I would hope that the unified segmentation of SPM5 would do a slightly better
job. Setting up the lesion masking is not done so elegantly in SPM5 though,
so you'll have to set up one segmentation at a time.
There are probably some segmentation approaches that handle lesions, but no
freely available implementations that I know of, other than EMS, which
requires an old SPM version.