 |
 |
Hello, I believe the apriori probability masks came from the ICBM probability atlas, and there are a few pages describing its derivation on the LONI website. http://www.loni.ucla.edu/ICBM/ICBM_Probabilistic.html http://www.loni.ucla.edu/ICBM/ICBM_TissueProb.html Also, from spm_segment: " The template image, and a-priori likelihood images are modified versions of those kindly supplied by Alan Evans, MNI, Canada (ICBM, NIH P-20 project, Principal Investigator John Mazziotta)." They say that 452 T1 scans were aligned into the atlas space, and segmented into grey matter, white matter, and csf. Given the high number of subjects, I would really be surprised if the scans were all acquired at the same field strength, with the same parameters, or on the same machine. In reading your concerns, it seems to me that the atlas would remain valid for classifying data from any scanner/field strength combination, and that the accuracy of segmentation would improve with increased CNR in a T1-weighted image. I say this for two reasons- first, the tissues were segmented and averaged together, and thus the apriori images do not reflect the specific MRI intensities and contrasts, but rather, the actual probabilities of a voxel belonging to a tissue class. Second, the segmentation algorithm is not dependent on the tissues being a certain intensity. However, your concern that the atlas may not match certain subject populations is overall I believe a more valid one. However, I would expect that it would match well enough to give a decent segmentation. Given the variability of different brains, it seems as if it would be difficult to make a better template from a small sample of patients (or normals who differ demographically) compared to the template constructed from such a large population. Ken ---------------------------------------------------- Ken Roberts Woldorff Laboratory Center for Cognitive Neuroscience, Duke University (919) 668-1334 ----------------------------------------------------