Four of 104 public SEEG excerpts had changing signals with runs of 31 to 340 equal samples at 512 Hz

We tested how the duration of an exact flat signal check changes its result. We used 104 saved excerpts of one second from 35 SEEG records and 24 people in OpenNeuro ds005624 version 1.0.0. The primary set had 14,252 finite contact signals that changed within their saved second. At a nominal duration of 3.90625 ms, 121,095 of 9,174,872 short windows were flat. At 15.625 ms, four changing contact seconds still had flat windows. Their longest runs had 31, 114, 252, and 340 equal samples at 512 Hz. Three runs touched an excerpt boundary, so their full source lengths can be greater. All tested durations through 500 ms had a nonzero count. Direct sliding-window checks matched all primary counts. Nine fresh source ranges matched the saved raw checksums. The package preserves raw witnesses, source metadata, fixed byte ranges, checksums, code, tables, and two figures. These are results on stored signals in sampled excerpts. They do not identify neural silence or measure decoder accuracy or stimulation effects.

SEEG · signal quality · flat signal detection · brain-computer interface · window duration · run length · OpenNeuro