Strong SEEG responses had 48 percent mean agreement within 10 ms for 20–300 ms peaks from unused pulse groups in five patients

We tested peak timing in public OpenNeuro SEEG dataset ds004457, version 1.0.2, from five patients with epilepsy. Four pulses selected the top 10 percent of eligible response pairs by energy within each patient. Two unused groups of three pulses gave separate absolute peak times in the 20–300 ms window. We tested 1,080 pairs and selected 111. Selected responses had 48.1 percent mean agreement within 10 ms. Omitted responses had 16.9 percent agreement. Each patient had the same weight. The difference was positive in all five patients. The exact two-sided test with patient sign flips gave p = 0.0625. This sample does not establish population reliability. All 2,100 pulse allocations gave 45.8 percent selected agreement. Removing the after-window mean changed agreement to 65.9 percent and changed the peak measurement. A discrete waveform proof shows that equal strong energy and correlation near one do not guarantee small peak error. It also gives a sufficient peak margin bound under a known maximum waveform error. The package preserves inputs, full code, source hashes, raw retrieval instructions, all pair and split results, figures, and source and calculation checks. The measured peak is not response onset or a validated conduction delay.

SEEG · stereoelectroencephalography · electrical stimulation · intracranial EEG · peak timing · response repeatability · waveform analysis · OpenNeuro · peak agreement · signal selection