search

spinicist / QUIT

A set of tools for processing Quantitative MR Images
Mozilla Public License 2.0
59 stars 20 forks source link

Clarification on MP2RAGE Implementaiton/Usage #37

Closed gdevenyi closed 5 months ago

gdevenyi commented 6 months ago

Reading over the MP2RAGE details: https://quit.readthedocs.io/en/latest/Docs/Relaxometry.html#qi-mp2rage

Its not clear to me what inputs are expected. Under our standard Siemens sequence we get out INV1, INV2 and UNI.

I'd like to get out a robust (denoised) T1-weighted image and the T1 map, can this function do this, or is it expecting some sort of raw data?

spinicist commented 6 months ago

Hello,

It expects the complex-valued INV1 and INV2 images in a single nifti as sequential volumes/timepoints. See here for where those images are then extracted back into separate volumes: https://github.com/spinicist/QUIT/blob/master/Source/Relaxometry/qimp2rage.cpp#L88

It will then output the UNI image and the T1 map as separate files.

gdevenyi commented 5 months ago

Okay, I was thinking of the robustcombination ( https://github.com/marcpabst/mp2rage-denoise/blob/master/mp2rage_denoise.py ) method for already-reconstructed UNIs with non-complex INV1 and INV2 images already processed by the scanner. Thanks.

spinicist commented 5 months ago

qi mp2rage already has the robust combination method implemented (and it's documented https://quit.readthedocs.io/en/latest/Docs/Relaxometry.html#qi-mp2rage !) with full control over the regularization parameter. Hence it needs the raw data as input.