Add duration column to _scans.tsv

[yarikoptic]

To accompany

• [ENH] Add optional 'duration' column to _scans.tsv bids-specification#2508

Add the optional duration column (wallclock seconds from first to end of last acquired sample) to three existing datasets and one new scans.tsv:

• synthetic (MRI, 5 sub × 2 ses): T1w=240s, nback-run-01/02=720s (=288×TR), rest=600s (=240×TR). All durations are multiples of TR=2.5s and fit within the acq_time gap to the next scan (no acquisition overlaps).

• ds000117 (MEG+EEG): duration=450s (=7.5 min) for all 6 runs per subject (16 subjects), consistent with the TaskDescription in the sidecar JSON ("six, 7.5 minute runs"). Verified no overlap: minimum start-to-start gap across all 16 subjects is 556s (sub-16 run-01→run-02), leaving a 106s residual.

• ds000246 (MEG, CTF): duration=360s for both AEF runs (matching RecordingDuration=360 in each run sidecar); duration=30s for the emptyroom noise run (RecordingDuration=30).

• ds000248 (MEG, Elekta/single-subject): duration=277.71s for the audiovisual run (RecordingDuration≈277.71 in sidecar); duration=110.0s for the emptyroom (RecordingDuration≈109.999).

• volume_timing (fMRI timing variants, new scans.tsv): durations derived from VolumeTiming + per-frame duration metadata in each sidecar. The acq-deprecated run (VolumeTiming 0–14s, AcquisitionDuration=1.0s per frame) gets duration=15.0s, consistent with the spec rule that the BOLD+VolumeTiming exclusion applies (AcquisitionDuration was used there as a per-frame measure, so the ScansAcquisitionDurationConsistency check is correctly not triggered).

• qmri_qsm (MRI with _part-, new scans.tsv): both part-mag and part-phase entries carry identical acq_time and duration (480s), illustrating that split-storage files of the same acquisition share temporal metadata.

• BIDS Specification PR: [ENH] Add optional 'duration' column to _scans.tsv bids-specification#2508

• BIDS Validator PR: feat: Add associations.scans loader bids-validator#449

• TODOs:

  • Drop TEMP commit after verying that we trigger CI to fail with failing data

Co-Authored-By: Claude Code 2.1.263 / Claude Sonnet 4.6 noreply@anthropic.com

[yarikoptic]

hm , TEMP is there and validator is still happy -- likely need to tune up the "breakage" example