Matched SEEG windows gave an uncertain 21 percent after-pulse increase in the spread-to-nearby forecast error ratio in eight patients

We tested whether three time-matched spread calibration records have a larger relative forecast error advantage for nearby records after a pulse than before it. The study used eight patients, 86 stimulation sites, and 886 remote bipolar readouts from OpenNeuro ds004696 version 1.0.1. Both 80 ms windows contained 164 samples. The equal-patient mean after-minus-before log contrast was 0.187818, an estimated 20.7% increase in the geometric spread-to-nearby error ratio. Its descriptive 95% whole-patient bootstrap interval was -9.2% to 79.4%; the exact two-sided sign-flip reference value was 0.46875, and four patients had positive contrasts while four had negative contrasts. The primary result leaves a larger relative after-pulse nearby-record advantage unresolved; it does not establish equal advantages. Patient 02 had the largest contrast; excluding that patient reduced the estimate to 1.1%. Fixed checks gave 39.31% for an earlier before-pulse window and 1.20% for regular first-five intervals. These overlapping secondary checks do not identify a cause. Fresh extraction from all eight complete public source files reproduced the parent after-pulse products and reported errors. The contribution fills a matched-duration control gap in the same clinical cohort, not an independent cohort confirmation. All physical pulses were delivered. The work does not establish a neural mechanism, an optimal stimulation schedule, or live interface performance.

stereoelectroencephalography · electrical stimulation · forecast error · calibration records · paired comparison · brain-computer interface