The variation fraction was higher at 50 to 250 ms than at 20 to 50 ms in eight-pulse SEEG maps from five patients

We tested whether across-pulse variation formed a different fraction of early and late mean waveform energy in public SEEG. The data were OpenNeuro ds004457 version 1.0.2: five epilepsy patients, nine bipolar recording pairs, and 322 stimulation-to-recording measurements. Each measurement used its first eight eligible pulses in both windows. Q was squared mean waveform voltage averaged over time. B was across-pulse sample variance averaged over time and divided by eight. For each patient, the variation fraction was sum(B)/sum(Q). Its mean was 8.78% at 20 to 50 ms and 20.94% at 50 to 250 ms. All five patient differences were positive. The mean difference was 12.16 percentage points; the descriptive 95% t interval was -3.37 to 27.70 points; the exact two-sided patient sign-flip value was 0.0625. A 50 to 80 ms late window gave positive differences in only three patients. A pre-stimulation control also gave a positive mean difference. These checks limit a neural explanation. The variation term is not a measured physiological noise fraction. The correction U=Q-B is established. It estimates stable mean waveform energy when errors from different pulses have zero covariance. This study adds a fixed-count window comparison, not a new general estimator or an independent replication. Fresh downloads of all 54 selected raw files reproduced every saved array element. The package preserves epochs, one complete raw pair, source hashes, code, results, figures, a local plan, all fixed checks, and a marked secondary current check.

stereoelectroencephalography · SEEG · electrical stimulation · cross-pulse variation · waveform energy · variance fraction · ventral temporal cortex · repeated measurements