Currently heuristics are saved within a private repository. This requires admins to create and maintain custom heuristics, while also making the conversion a bit of a black-box to the end-user. If we want to shift onus of maintaining studies more to the end-user, there can be a two-stage implementation to change the current workflow.
As a stop-gap, move the heuristics to a public repo, behind a protected branch. This will allow users to make changes or add new heuristics via pull request, lessening the burden on admin. Admin will still need to be accept and merge in pull requests. Heuristics are pulled automatically once a day, or alternatively, admins can trigger a heuristic update via the admin page.
Refactor the use of heuristics similar to the .bidsignore contents. That is, add a column to the study table in the database saving a default heuristic (cfmm_base?). Under the study config, the user can then edit this study config and each time tar2bids is called, the heuristic is written to a temporary file.
With (2), the user will have to maintain their own heuristics. Ideally, there would be a guide for this (on CBS Discourse?). Would need to think around how users can get access to the heudiconv generated files to aid with the mapping if necessary. Alternatively, could users get access to DICOM values from the DICOM server?
Currently heuristics are saved within a private repository. This requires admins to create and maintain custom heuristics, while also making the conversion a bit of a black-box to the end-user. If we want to shift onus of maintaining studies more to the end-user, there can be a two-stage implementation to change the current workflow.
.bidsignore
contents. That is, add a column to the study table in the database saving a default heuristic (cfmm_base
?). Under the study config, the user can then edit this study config and each timetar2bids
is called, the heuristic is written to a temporary file.With (2), the user will have to maintain their own heuristics. Ideally, there would be a guide for this (on CBS Discourse?). Would need to think around how users can get access to the heudiconv generated files to aid with the mapping if necessary. Alternatively, could users get access to DICOM values from the DICOM server?