Ghislain Vaillant
Ghislain Vaillant
As far as user experience is concerned, perhaps a simpler solution would be to just discard processing of sessions not containing the targeted modality instead of crashing?
> When launching a pipeline related to a specific modality (T1w, PET), the pipeline will crash if one session of the BIDS does not have this modality available. Besides, even...
Which raises the following question: in which precision do we want our NifTi stored in the end? Aka, the file size vs dynamic range dilemma.
Back in my PhD days, we were using short integers (`int16` / `uint16`). I don't know the rationales for this choice at the time.
As discussed with @nburgos, short integers should be enough for storing MRI and PET modalities with a reasonable dynamic range.
I like the split of the documentation pipeline to separate build and publish stages. I also like the replacement of `scp` with proper Jenkins constructs such as [sshPublisher](https://www.jenkins.io/doc/pipeline/steps/publish-over-ssh). Imo, the...
Do we really gain something from having separate `build` and `deploy`stages coupled by stashing and unstashing? Since the same node is capable of building and deploying, it sounds to me...
> And that let us be more flexible if we add more agents in the future I agree that separating those stages allow for more recycling of nodes for other...
Good point. We'd need an extra version check for FreeSurfer 6 when using the FastSurfer option.
FreeSurfer version 7.2 is now available and has no breaking changes with the 7.1 series according to the release notes.