ASL-T1 registration

[caugolm]

I've run a number of images through ASLPrep and things are generally looking all right (ie, pretty much every scan is getting output, most CBF values look biological, etc). However, in looking through the html files, it appears as though somewhere between 30 and 40% of the data has a moderate to severe issue with the ASL-T1 registration. I've saved all working directory data and started to try to hunt down a way to manually fix the registrations and then re-start the processing, but is there an intended way to manually fix and rerun ASLPrep using the fixed registrations? Thanks for any help!

[caugolm]

@tsalo After quite a bit of testing, it seems that this is due to the choice of the "aslref" image and the particular pulse sequence, which is a 3D stack-of-spirals that has some "ringing"-like artifacts in individual volumes that is removed upon the label control subtraction and time series averaging. This ringing affects the detection of the brain boundary, which creates problems for the ASL brain masking and for the aslref-T1 FLIRT registration. But, if I use the output mean-cbf image and re-run FLIRT between that and the T1 and then things look much better (and on top of that I can apply BBR and it seems to consistently perform well). I feel like a potentially important caveat is that I'm still using ASLPrep 0.2.8.

[tsalo]

@caugolm, I'm not surehow I missed this, but thanks for the followup post. Hopefully addressing #343 will also address this problem.