BF: relax filename regex — non-numeric ses, optional run, extra BIDS entities

[yarikoptic]

@yarikoptic disclaimer -- this PR is solely AI driven but does relate also to

• Consider an "extended" set of entities beyond run #13
• NIDM extraction: 51/219 datasets failed in the 2025-05-16 run OpenNeuroDerivatives-NIDM#7

and on a brief review looks good to me.

Summary

The current fname_re in summarize_confounds fails on many real BIDS/OpenNeuro derivative filenames, causing the script to hard-exit on the first confound file. Discovered while running the pipeline in ReproNim/OpenNeuroDerivatives-NIDM — 28 of 219 datasets on 2025-05-16 aborted at line 105 with unsupported filename: sub-... (see the tracking issue).

Three real-world filename shapes the current sub-(\w+)(_ses-\d+)?_task-(\w+)_run-(\d+) cannot handle:

1. Non-numeric session IDs — ses-timepoint2, ses-test, ses-func11, ses-T1, ses-pre (5 datasets)
2. No _run- entity — single-run acquisitions like sub-01_task-rest_desc-confounds_timeseries.tsv (17 datasets)
3. Extra BIDS entities between task- and run- — e.g. sub-13_task-MemorySpan_acq-multiband_run-01_... (ds000237)

The fix

fname_re = re.compile(
    r'sub-(?P<subject_id>[^_]+)'
    r'(?:_ses-(?P<ses>[^_]+))?'
    r'_task-(?P<task>[^_]+)'
    r'(?:_(?!run-)[a-z]+-[^_]+)*'  # optional entities: acq, rec, dir, echo, space, ...
    r'(?:_run-(?P<run>[^_]+))?'
)

• Named groups (rather than positional) so re_groups and the regex stay in sync.
• [^_]+ per entity — permissive on ID shape but still anchored to BIDS entity boundaries.
• The (?:_(?!run-)[a-z]+-[^_]+)* middle segment allows any number of BIDS entities to sit between task- and run-, with a negative lookahead on run- so it doesn't swallow that entity itself.
• Extraction switched from dict(zip(re_groups, m.groups())) to {k: m.groupdict().get(k) for k in re_groups} — column ordering now follows re_groups, not regex group order.

Verification

Tested against the 28 filenames observed to fail in the OpenNeuroDerivatives-NIDM May-2025 run (.duct logs). All 28 now match and yield the expected (subject_id, ses, task, run) groups — including edge cases like sub-MSC10, sub-rid000041, sub-13_task-...acq-multiband_run-01, sub-01_task-rest (no run), sub-080_ses-T1.

Commits

• b0010a1 BF: relax filename regex — non-numeric ses, optional run, extra BIDS entities

Test plan

• All 28 previously-failing filenames now match with correct named groups
• Numeric-only filenames (previously supported) still match
• Downstream summarize_confounds CSV output still validates against metadata/summarize_confounds_fmriprep_dictionary.csv on a real dataset (reviewer to spot-check)

[gemini-code-assist]

Code Review

This pull request refactors the regular expression in summarize_confounds to use named groups and handle optional BIDS entities and runs, updating the dictionary merging logic accordingly. The review feedback suggests enhancing the regex to support alphanumeric BIDS entity keys for better robustness, and simplifying the dictionary update by directly using m.groupdict().

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

To make the regex more robust against custom or non-standard BIDS entity keys (which may contain uppercase letters or numbers, e.g., Acq, rec2), we should allow alphanumeric characters [a-zA-Z0-9] instead of just lowercase letters [a-z] for the entity keys.

Suggested change

	r'(?:_(?!run-)[a-z]+-[^_]+)*' # optional BIDS entities between task and run (acq, rec, dir, echo, space, ...)
	r'(?:_(?!run-)[a-zA-Z0-9]+-[^_]+)*' # optional BIDS entities between task and run (acq, rec, dir, echo, space, ...)

[gemini-code-assist]

Since fname_re defines exactly the named groups corresponding to re_groups, m.groupdict() will return a dictionary containing exactly those keys with their matched values (or None if they didn't match). We can simplify the update of c_sum by directly passing m.groupdict() instead of using a dictionary comprehension.

Suggested change

	c_sum.update({k: m.groupdict().get(k) for k in re_groups})
	c_sum.update(m.groupdict())