Hi Chintan - What makes you think that it's bedpostx_gpu? Does the problem go away when you use regular bedpostx?
It's hard to guess what's going on. If you send me a subject that works and a subject that doesn't (zip all tracula dirs of these subjects) here, I'll take a look: https://gate.nmr.mgh.harvard.edu/filedrop2/
Best, a.y ________________________________ From: freesurfer-bounces@nmr.mgh.harvard.edu [freesurfer-bounces@nmr.mgh.harvard.edu] on behalf of Mehta, Chintan [chintan.mehta@yale.edu] Sent: Monday, October 23, 2017 3:24 PM To: freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] Tracula --- Volume and path length are equal
Dear FreeSurfer users,
I am using FreeSurfer 6.0 and Tracula 6.0. For many subjects that I'm processing, the volume and path length are equal to each other. For ~60% of subjects, the "pathstats.overall.txt" file shows something like:
Count 1500 Volume 34 Len_Min 34 Len_Max 34 Len_Avg 34 Len_Center 34 AD_Avg 0.00104677 AD_Avg_Weight 0.00104677 AD_Avg_Center 0.00104677 RD_Avg 0.000594649 RD_Avg_Weight 0.000594649 RD_Avg_Center 0.000594649 MD_Avg 0.000745357 MD_Avg_Weight 0.000745357 MD_Avg_Center 0.000745357 FA_Avg 0.364984 FA_Avg_Weight 0.364984 FA_Avg_Center 0.364984
This output suggests that only a single pathway was computed for a subject with output like above (since Len_Min = Len_Max = Len_Avg = Len_Center); yet the count number suggests that 1,500 paths were generated.
This obviously seems problematic. I have two questions:
1) Does the problem originate during the fitting of the ball-and-stick models at each voxel? (I used n=3 sticks with bedpostx). If so, how can I rerun this process to ensure this problem does not repeat?
2) Is it possible that each probabilistic track was generated with the same random seed? If so, would this problem be associated with the ball-and-stick model step (with bedpostx_gpu) or pathway fitting step (-path)?
3) Are the anisotropy measures (such as FA_Avg, MD_Avg_Weight, etc) independent of this problem? Or do they also require different probabilistic paths be generated?
I suspect this issue may have arisen due to my using "bedpostx_gpu" to fit the ball-and-stick models (default setting).
Thank you much in advance.
Best,
Chintan