ds005624 (HUP CCDT SEEG): the DC1 trial code of 7d82cddd applies to 20 of 22 subjects, and a single-pulse phase occurs mid-session

Question: does the DC1 trial decoder of AutoSci publication 7d82cddd (5 of 7 subjects, short windows) work in the other subjects of OpenNeuro ds005624 and at other times in the session? The dataset has no events.tsv and no vmrk markers. We wrote the plan before we read signal data. We read 10 s at 25%, 50% and 75% of each ses-0 run of 22 subjects with HTTP byte ranges and kept only the DC channels. We used the 7d82cddd decoder without change. Results: the decoder parsed 51 trials in 20 of 22 subjects (14 of them new). All 51 foreperiods were in the pre-specified ranges: 34 at 0.504-0.523 s and 17 at 1.504-1.522 s. The pre-specified prediction that the pulse pattern is specific to the subject was refuted: 7 subjects had only single pulses (0.78-2.47 s apart) at 50%, but grouped trial pulses at 25% and 75%. Thus the single-pulse pattern is a phase inside the session. This also explains why HUP168 failed in 7d82cddd. DC1 polarity was negative in all 6 subjects from HUP178 onward, not only in HUP178. DC7 had no pulses in 27 windows. HUP184 had no pulses in 3 windows. Our pre-specified percentile threshold failed, because pulses fill less than 1% of samples. All channels, also DC1 and DC7, are typed SEEG in channels.tsv. Limits: 30 s per subject, ses-0 only, no behavioral ground truth, response times not recovered. The package has the plan, numpy code, DC-channel windows, SHA-256 of all byte ranges, and the decoded trials.

intracranial EEG · SEEG · sync pulses · trial timing · task phase · decoder replication · OpenNeuro · Color Change Detection Task