Separate five-pulse profiles had higher mean SEEG spatial error than a shared profile at 20–500 ms in seven patient sessions

We tested whether a separate spatial SEEG profile from five pulses at each stimulation site predicted final-five-pulse profiles better than one shared profile. We used OpenNeuro ds004696 version 1.0.1. Fixed rules retained 57 sites in 20 groups from seven patient sessions. Each group used one stimulation shaft, one current, and a common recording panel. Shaft energy normalization removed one common response size factor. Separate profiles had mean squared Hellinger error 0.02672; the shared profile had error 0.02313. Mean error reduction was -0.00358, with a descriptive 95% patient bootstrap interval of [-0.01169, 0.00451] and exact two-sided patient sign-flip p = 0.5469. The after-minus-before contrast was 0.00593, with p = 0.2188. The fixed joint evidence rule failed. Pulse order limits the claim to an offline comparison. This result does not establish equal neural responses or independent neural control. The package includes a reusable profile method, unchanged exact pulse products, all profiles and scores, three figures, a local plan, independent arithmetic checks, source provenance, and reproduction instructions. Fresh audits of two complete source derivatives passed 2,496 direct energy checks. All 52 source snapshot checks passed. A repeat run gave identical numerical files.

stereoelectroencephalography · electrical stimulation · spatial response profiles · squared Hellinger error · brain-computer interfaces · patient bootstrap · sign-flip test · offline prediction