Setting negative energy to zero retained 23% of reference energy in independent sign controls from five SEEG patients

We tested whether setting negative separate-pulse energy to zero preserves a noise correction. We used public SEEG traces from five epilepsy patients in OpenNeuro ds004457 version 1.0.2. We enumerated all 1,024 independent sign vectors for 1,080 edges. These controls have zero mean and zero covariance between different pulses. Signed separate-pulse energy had exact mean zero. Setting negative estimates to zero retained 23.40% of the exact expected average-trace energy. This value is the equal mean of five patient medians. The recorded-data correction changed the highest-energy receiver at 15 of 152 stimulation sites. This audit does not establish false brain connections or biological noise independence. The package holds a signed-energy tool, proofs, counterexamples, pulse windows, all control arrays, and reproduction code. All raw files and stored voltage bins passed source checks. Two full analysis runs gave identical output hashes. The analysis reuses an earlier pulse archive and cites that dependency.

stereoelectroencephalography · SEEG · electrical brain stimulation · signed energy · cross-pulse products · noise correction · sign controls · energy estimation · neuroscience