pymef 1.4.8 decoded non-zero data for only one stimulation site in each of four ds004977 SEEG runs

We read four SEEG cortico-cortical evoked potential runs from OpenNeuro ds004977 v1.2.0 with pymef 1.4.8. In every run, all good SEEG channels had one continuous block of non-zero samples, 255,394 to 255,540 samples long (about 53.2 s at 4800 Hz); all other samples were exactly zero. The non-zero fraction was 7.0% to 14.0% per run. Only one stimulation site in each run had pulse epochs with non-zero data: 4 of 38 sites and 48 of 455 good pulses in total. In those epochs, median channel standard deviation was 1.3e5 to 4.7e5 in the stated unit, microvolts, which is not physiological; a scale or decode error is possible. Local file sizes matched the listed S3 object sizes, but a second MEF3 reader was not tested, so the cause may be the files or pymef 1.4.8. We did not perform the pre-registered comparison of common average reference with adjusted common average reference because 86% to 93% of sites had no non-zero data in our read. An aborted single-site output is included and is not a CCEP result. Users decoding ds004977 with pymef should inspect the decoded data before analysis.

SEEG · cortico-cortical evoked potentials · OpenNeuro · MEF3 · pymef · data decoding · common average reference