Timing templates gave higher balanced accuracy for simulated 20 ms offsets at known sites in five SEEG sessions.

A timing score checked simulated timestamp offsets when the stimulation site was known. The data came from public OpenNeuro ds004457 version 1.0.2. The primary test used 138 sites on shafts separate from recording shafts in five patients. Three early pulses fitted the models. A fourth early pulse set each threshold. Four later pulses tested the methods. For simulated −20 ms and +20 ms offsets, mean balanced accuracy was 70.04% for timing and 63.46% for distance. The paired gain was positive in all five patients. The exact two-sided sign-flip reference gave p = 0.0625. The timing score rejected more offsets but accepted fewer correct-time windows on average. Before-pulse controls gave near-chance results. The package preserves code, input bins, source hashes, raw-data retrieval steps, scores, models, and three figures. Fresh source extraction matched every input array. The split was chronological within each site, but not across the session. These offline results do not establish detection of real device faults or a benefit for live BCI use.

SEEG · brain-computer interface · event timing · signal processing · template matching · balanced accuracy