A linear pulse trend did not establish higher test SEEG voltage reduction at 3 to 15 ms in 18 records with 16 fit pulses

We tested whether a linear trend across calibration pulses improves subtraction of an alternating SEEG voltage template. We used one selected target record from each of 18 people in OpenNeuro ds005169 version 1.0.0. Each local fit used 16 pulses. Each test used the next 32 pulses and offline detected edges. At 3 to 15 ms, the trend fit increased mean reduction by 0.0797 percentage points relative to a balanced fit. The 95 percent person bootstrap interval was -0.0626 to 0.2578 points; the exact sign-flip reference was 0.4167. The test did not establish higher mean reduction. Both fits slightly increased mean late-window voltage variation. Under a fixed additive waveform model with independent pulse errors of equal covariance, trend rejection removes linear drift bias and increases template noise variance by 1.19 percent for 16 equal pulse intervals. We checked the proof identities and all 504 scores with independent least-squares fits. All 18 complete source files matched pinned hashes, and all rebuilt arrays matched the archived inputs. The package preserves compact data, fitted templates, result tables, three figures, source metadata, code, and reproduction instructions. This result measures voltage variation. It does not establish neural recovery or better decoder performance.

stereoelectroencephalography · SEEG · electrical stimulation · pulse template · signal processing · voltage variation · linear trend · template subtraction