Hey everyone. I was wondering if anyone had any insight into why I might be getting odd data after using fslmaths and fslstats. Here's what I've done so far. Ultimately I'm trying to get the average FA and volume of the resulting pathway from probtrackx:
1) Ran probtrackx using multiple masks (had two seeding points in structural space) for each hemisphere using default values
2) Using ApplyXFM_gui, moved dti_FA to structural space (used diff2str.mat as transformation matrix) - resulting file is called dti_FA_str
3) Binarized fdt_paths and then multiplied it by dti_FA_str using this command (note when I originally got advice on this way back when I was told to use avwmaths_32R for this step but that doesn't exist anymore so I used fslmaths)
fslmaths fdt_paths -thr 50 -bin -mul dti_FA_str fdt_paths_values
4) Retrieved average FA and volume using this command
fslstats fdt_paths_values -M -V
So here's my issue. This appears to work, although I've run into some subjects where the output for average FA is 1.000000 while still having a reasonable volume value. I opened a subject's dti_FA_str and added fdt_paths and fdt_paths_values. There are 1) a number of voxels that have intensities greater than 50 that are not included in fdt_paths_values and 2) all the voxels in fdt_paths_values have intensities of 1 (even though dti_FA_str has clear values consistent with white matter FA). I looked at the same files in a subject where I got actual FA values and fdt_paths_values looks much less patchy and appears to encompass all voxels greater than 50. What's odd is that for all but one subject, this seems to only be happening in one hemisphere. For instance, the last subject I pulled, the data was fine for the left hemisphere but FA=1 came out for the right hemisphere.
I thought maybe somehow fdt_paths_values was getting binarized, so I tried running fslmaths first with -thr and -bin and then running it again with -mul, but I still get the same thing. All the voxels in fdt_paths_values still had a value of 1.
Does anyone have any idea what's going on here and how to fix it? Also, since I'm writing, I may as well check and see that people still think using a threshold of 50/5000 is acceptable for this analysis (connecting two masks that are known to be connected).
Thanks in advance for your help. Sorry for the long e-mail...
Kristen
|