The image extent check does not select a task time column in 27 ds000030 control runs.
An audit of 126 selected control runs from OpenNeuro ds000030 version 1.0.0 compared task event times with time limits calculated from raw image headers. All headers declared 184 volumes at 2 s, giving an image extent of [0, 368) s. Every active onset in both task columns fell within that extent. The columns differed in 27 runs, so the image extent check did not select a task time column in those runs. The stored onset differences were 0 s in 99 runs, +2 s in 26, and +4 s in one; the source README describes an 8 s correction. The audit does not measure the mapping between the task clock and the first saved volume and does not show that the correction is wrong. Hardware timing error, the meaning of trigger_time, and the cause of the 590 s scan-time value for sub-10325 remain unresolved. Results apply to the selected raw files and do not test image signals, clinical groups, or biological causes.
fMRI · task timing · image headers · event onset · OpenNeuro · ds000030 · NIfTI · BIDS · data audit · reproducibility
References
4- usesTwo methods agree on cue predictor changes to 1e-9 in 126 ds000030 control runs. sci:sha256:389d2890e7a3886f007771f9bfb71758ffd5cd5f39b21c32d34c4658cbe584dc
The file list, source checksums, and selected run list gave the same 126 runs for this header audit.
- usesdoi:10.18112/openneuro.ds000030.v1.0.0
The audited task tables, participant selection, JSON metadata, and pinned image prefixes are from this exact dataset version.
- supportsdoi:10.1038/sdata.2016.110
The dataset report documents the scanner task and the source description of its event-time correction.
- doi:10.7554/eLife.88652
This related image analysis is discussed as context for its use of GO and STOP task times; the present audit does not reproduce its image analysis.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesAll seven recorded delay differences of more than 50 ms occur in ds000030 1.0.0 files without `trigger_time`.
The read metadata inventory identifies the trigger_time field and missing values in 126 control runs. That scoped evidence gap helped specify the complete 266-run comparison of field presence with large delay differences. No image-header calculation or demographic selection is reused.