search

pyushkevich / greedy

Very fast greedy diffeomorphic registration code
GNU General Public License v3.0
59 stars 16 forks source link

Can it work on Sinogram level for motion correction? #39

Closed WenhongLan closed 1 year ago

WenhongLan commented 1 year ago

Hello all,

I tried to use greedy for image registration on sinogram level. my sinogram size is 520 x 50 x 11559 x 35 (int8). I configure the affine and deformable registration. I am following the quick start instruction of Greedy. and greedy can run for the lowest and intermediate resolution. But when working on full resolution, the greedy will be interrupted after finishing the intermediate resolution iterations (on Affine mode, no affine.mat output).

I am considering whether it is the hardware or the software restriction. May I ask does any hard restriction on the application level?

Additionally my hardware is Xeon Gold 6242R with 768 GB RAM.

Best regards Wenhong Lan

pyushkevich commented 1 year ago

Hi, I've never tried it on anything that size. In principle you should have enough memory. Have you tried subsampling the image along the large direction, maybe to make it 510x50x1160x35 and then trying?

On Tue, Jun 6, 2023 at 10:21 AM wenlan @.***> wrote:

Hello all,

I tried to use greedy for image registration on sinogram level. my sinogram size is 520 x 50 x 11559 x 35 (int8). I configure the affine and deformable registration. I am following the quick start instruction of Greedy. and greedy can run for the lowest and intermediate resolution. But when working on full resolution, the greedy will be interrupted after finishing the intermediate resolution iterations (on Affine mode, no affine.mat output).

I am considering whether it is the hardware or the software restriction. May I ask does any hard restriction on the application level?

Additionally my hardware is Xeon Gold 6242R with 768 GB RAM.

Best regards Wenhong Lan

— Reply to this email directly, view it on GitHub https://github.com/pyushkevich/greedy/issues/39, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJPEWYXA4QUJ6AD6SAEAL3XJ3SARANCNFSM6AAAAAAY4BNRIY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

WenhongLan commented 1 year ago

Many thanks for you comments. I am testing with the down-sampling data, although we do not expect any down-sampling on the data.

Besides, we also test the dataset I mentioned with greedy in another workstation with 4TB RAM. And it still does not have outcome. But greedy works, at least from the log file. And I am considering whether the problems on the I/O function used by Greedy, such as the old version .mat format with the maximum size restriction.