Dear list, especially Bruce and Jon,
We are starting to acquire images on the local 7T Philips scanner and would like to run them through FS. We did some pilot runs following what's described on http://surfer.nmr.mgh.harvard.edu/fswiki/SubmillimeterRecon and in http://www.sciencedirect.com/science/article/pii/S105381191730808X. That took approximately 42 hours. It looked reasonable but I estimate from the pilot run that some manual edits will be in order. This is quite tedious to do if the turnaround time is 2 days and we are looking at quite a few number of patients we want to run.
Would we actually get any speedup if we utilize the -use-gpu flag (we don't have access to a gpu yet, but are getting one soon). I looked at the speedup factors and especially mris-inflate doesn't speed up too much (https://surfer.nmr.mgh.harvard.edu/fswiki/CUDADevelopersGuide) .
Kind regards,
Mel
Hi Mel
most of the time is usally in mri_ca_register, which does get a big benefit from GPU acceleration. mris_inflate should only take a couple of min, so I wouldn't worry about it.
Not many recons take 42 hours any more though! That probably means you had a lot of largish defects. Are you using the 7T acquisition protocol that Jon has recommended?
cheers Bruce
On Mon, 13 Nov 2017, Melanie Ganz wrote:
Dear list, especially Bruce and Jon,
We are starting to acquire images on the local 7T Philips scanner and would like to run them through FS. We did some pilot runs following what’s described on http://surfer.nmr.mgh.harvard.edu/fswiki/SubmillimeterRecon and in http://www.sciencedirect.com/science/article/pii/S105381191730808X. That took approximately 42 hours. It looked reasonable but I estimate from the pilot run that some manual edits will be in order. This is quite tedious to do if the turnaround time is 2 days and we are looking at quite a few number of patients we want to run.
Would we actually get any speedup if we utilize the -use-gpu flag (we don’t have access to a gpu yet, but are getting one soon). I looked at the speedup factors and especially mris-inflate doesn’t speed up too much (https://surfer.nmr.mgh.harvard.edu/fswiki/CUDADevelopersGuide) .
Kind regards,
Mel
Hi Mel, what resolution are your scans?
doug
On 11/13/2017 09:59 AM, Bruce Fischl wrote:
Hi Mel
most of the time is usally in mri_ca_register, which does get a big benefit from GPU acceleration. mris_inflate should only take a couple of min, so I wouldn't worry about it.
Not many recons take 42 hours any more though! That probably means you had a lot of largish defects. Are you using the 7T acquisition protocol that Jon has recommended?
cheers Bruce
On Mon, 13 Nov 2017, Melanie Ganz wrote:
Dear list, especially Bruce and Jon,
We are starting to acquire images on the local 7T Philips scanner and would like to run them through FS. We did some pilot runs following what’s described on http://surfer.nmr.mgh.harvard.edu/fswiki/SubmillimeterRecon and in http://www.sciencedirect.com/science/article/pii/S105381191730808X. That took approximately 42 hours. It looked reasonable but I estimate from the pilot run that some manual edits will be in order. This is quite tedious to do if the turnaround time is 2 days and we are looking at quite a few number of patients we want to run.
Would we actually get any speedup if we utilize the -use-gpu flag (we don’t have access to a gpu yet, but are getting one soon). I looked at the speedup factors and especially mris-inflate doesn’t speed up too much (https://surfer.nmr.mgh.harvard.edu/fswiki/CUDADevelopersGuide) .
Kind regards,
Mel
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Douglas N Greve -
Melanie Ganz