In a 409-subject SEEG audit, all 264 clear line peaks agreed with numeric BIDS values; 131 recordings were unclear

Question: Does the BIDS sidecar field PowerLineFrequency agree with the mains frequency in the signal of public SEEG data? Pipelines use this field to select notch filters. Method: For the first 5 SEEG subjects in each dataset in the archived census (29fbc0d0), and for all 254 SEEG subjects of ds004789, we read 10 s of all channels with HTTP range reads. We calculated a channel-median line score at 50 Hz and 60 Hz and their harmonics, and classified the line as 50 Hz, 60 Hz or unclear with thresholds fixed in a hashed plan before the data read. Results: 409 subjects in 35 datasets were readable. All 264 subjects with a clear line and a numeric sidecar value agreed, and 0 disagreed (12 at 50 Hz, 252 at 60 Hz). 131 recordings (32%) had no clear peak; this can be because of a notch filter before export. In ds004789 the unclear fraction varied by RAM site letter (D: 15 of 26; P: 3 of 23). ds004473, ds005398, ds005415 and ds005624 give “n/a”, and ds004819 has no sidecar; 13 of 15 tested subjects in ds004473, ds005415 and ds005624 have a clear 60 Hz line. Sidecar SamplingFrequency agreed with the file header in 408 of 409 subjects. Limits: one 10 s window and one file per subject; only 5 subjects per dataset except ds004789; an unclear result cannot show a wrong value. The package includes the code, plan hash, S3 key of every file read and per-subject table. This version corrects the count of 278 in e231b0de to 264.

SEEG · intracranial EEG · OpenNeuro · BIDS · power line frequency · mains interference · signal processing · data quality