Solution for running L2 analysis / preprocessing

[dariusvalevicius]

Currently the best practice for storing BIDS data seems to be having the dataset description, participants tsv, and a set of BIDS single subjects at the top level of the CBRAIN project. Batch processing can be done with a file list.

However, for some tools that do group analysis (e.g. population module in ExploreASL), the only way to get the correct input right now seems to be to copy all the BIDS subjects into a file collection, duplicating the data.

Is there a better way to do this or a solution we can come up with?

[MontrealSergiy]

Hi Darius, I believe at the moment a most economical solution is to have a dedicated group analysis ExporeASL tool that takes a cbrain list of subjects folder, participant tsv and other files and folders of the bids dataset (not sure exact ExporeASL requirements and conventions, some tool might have use of 'preprocessed' or 'derivative' folder or dataset). list could be named by the dataset_name

The boutiques input should be explicitly declared as list, otherwise you will enable parallel treatment of subjects, which is not appropriate for group analysis.

REMARK 1 not sure though how to select subset in an efficient manner, as standard bids is not super powerful in subj selection. it might be tool specific query language, or you can use a dedicated package

REMARK 2 You will use command like app_name $PWD output_dir group . I do not sure will it work smoothly, one issue is if you do not delete output_dir each rerun, bids validator might complain. Otherwise you can add a command line creating input subfolder and placing original subject and meta data (or links to them) there, in fact I believe Natacha is a module/property to do just that.

I vaguely remember for some reason Natacha and Pierre created boutiques several custom module(s)/propreties(s) to copy input data into a subfolder, or extract outputs from subfolder. though I am not even 100% sure does module integrator work well in the present of list inputs in boutiques, but if not it should be fixed.

REMARK 3 I am not sure is input the original bids dataset or some 'preprocessed'

[MontrealSergiy]

I think, in particular, the following 'dir_maker' Boutiques module can be used to simplify L2 analysis

https://github.com/aces/cbrain/pull/1267