 |
 |
> 1) About the input data, can I just directly feed the MRI scan to the > segmentation module? Sometimes, the input data may be very noisy, > should I first smooth the scan first before sending it to the segment > module. Is smoothing the scan a routine preprocessing step that a > user should do before doing segmentaiton? Just run it through and see if the result seems acceptable to you. If it looks like the segmentation worked well by eyeballing it, then you are good to go. If not, you have a few options. You CAN smooth it, but there are reasons it might not be the best idea. A Gaussian blur will smooth out the noise, but it will also smooth over the boundary between gray/white matter. Using another package that implements, say, a Kalman filter or other "adaptive" filtering scheme would be a preferable way to preprocess. But, probably the best way is to get more acceptable images. I don't know what your scanning parameters are, but there should be a way to get a high-quality whole brain anatomical image in not too much time. Have you asked your MR physicist to use an IR prepped T1 acquisition? > 2) About the output data, I noticed that the outputs on both the gray > matter and the white matter are not binary image. I wonder what does > different value for each voxel mean. In addition, how can I get a > binary mask of the gray matter and white matter? The values are the probability that each voxel is of a given tissue class. They should fall on the range [0 .. 1]. > 3) Does SPM5 support extracting cerebral white matter and cerebral > cortex from the white matter and gray matter respectively? No, it doesn't do a cortical surface extraction like FreeSurfer does. > Thanks, > Kai > ---------------------------------------------------- Ken Roberts Woldorff Laboratory Center for Cognitive Neuroscience, Duke University (919) 668-1334 ----------------------------------------------------