Exchanged later voltage changed a mean 43.1% of site labels before pulses in five SEEG sessions with a centered 30 Hz FIR filter

We tested whether a site reader followed the source of exchanged later voltage. We used public SEEG data from five patients in OpenNeuro ds004457, version 1.0.2. The cohort had 138 site instances and 552 separate test pulses. The reader used the interval from 80 to 20 ms before a pulse. A centered 513-coefficient 30 Hz FIR filter also used later voltage. We exchanged voltage after a past-only 15 ms pulse repair and kept the trained reader fixed. The main donor-minus-recipient accuracy difference was 1.80 percentage points. Its two-sided sign-flip reference value was 0.1875. Evidence for donor-site identification was weak. Exchanged later voltage changed a mean 43.10% of labels. Every patient had changed labels. The causal filter and the distant primary control window changed no labels. Fixed secondary checks changed labels but did not give a stable mean donor preference. The result establishes future-input dependence in the tested analysis. It does not show neural anticipation or live BCI benefit. The package has native voltage excerpts, figures, source hashes, models, exchange results, code, and reproduction instructions. Independent extraction and analysis rebuilds matched all saved arrays.

stereoelectroencephalography · SEEG · brain-computer interface · causal filtering · finite impulse response filter · future-input dependence · signal processing