Transferred SEEG templates did not establish better early mean voltage reduction than fits from two local pulses in 18 contact pairs

We tested SEEG template transfer across stimulation contacts in 18 people from OpenNeuro ds005169 version 1.0.0. The fixed sample has 36 records. Transfer used 16 source pulses and one unbounded signed gain per channel, fit from two target pulses at -1 to 2 ms. All methods were scored on target pulses 17 to 48. Early mean voltage reduction was 71.39% for transfer and 80.43% for a fit from two local pulses. The paired difference was -9.04 percentage points, with a 95% person bootstrap interval of -21.92 to 1.21 and an exact sign-flip p value of 0.1559. This test did not establish a transfer advantage. At 3 to 15 ms, transfer added variation in all 18 people. The median increase was 21.45%; one failure gave a residual-to-recorded power ratio of 149.50. Shape permutations and baseline fits separate waveform information from reduced fit noise. The package preserves extracted epochs, source hashes, code, individual results, three figures, and a proof under stated noise assumptions. All 36 full source hashes and 144 independent byte-range checks passed. A numerical replay matched every saved table and template array. Voltage reduction does not establish neural recovery or decoder benefit.

stereoelectroencephalography · intracranial EEG · neural interfaces · electrical stimulation · template transfer · signal processing · voltage reduction