[MINC-users] mean signal

[Andrew Janke]

Hi Ayman,

>  I need to be able to obtain the mean signal intensity of an ROI in a brain MR image.  I thought of using Display to select the ROI, then save the selection as .mnc, then running volume_stats on the volume with a mask of the ROI.  However, two problems are associated with that: When running volume_stats on the mask alone, I get dimensions of the mask that are different than when running the mask on top of the volume (e.g. The mask has 9 voxels, but when I run volume_stats on the volume and mask, the # of voxels is now 32!!!), my other problem is that I always get the error: WARNING: Dimensions of RR_ROI_2.mnc do not match those of 0.mnc. Using world coordinates.  What does this imply for my measurements?

As per what Mishkin said, Display tries to save space on your HDD (And
save a smaller mask). So this means that using masks in
mincmath/mincstats is generally a two-step process.

  1. create mask with Display

  2. resample the data to the file you want to use it with:

    $ mincresample -like T1.mnc -nearest_neighbour mask.mnc mask.resampled.mnc

  3. Then use the mask with mincstats (instead of volume_stats).

    $ mincstats -mask mask.resampled.mnc -mask_binvalue 1 T1.mnc


I would guess that you have the 22 voxels with what you are doing
above due to partial volume effects around the mask when you are
resampling. This should not happen with the above when you use the
-nearest_neighbour flag with mincresample.


a


-- 
Andrew Janke (a.janke at gmail.com || http://a.janke.googlepages.com/)
Canberra->Australia +61 (402) 700 883