Early RMS lowered late SEEG forecast error beyond pre-pulse RMS in linear models with five calibration pulses per condition in eight STReEF patients

We tested whether early voltage size improves a late voltage forecast beyond pre-pulse voltage size. We used public SEEG data from OpenNeuro STReEF ds005448, version 1.0.0. The sample had eight patients, 96 selected conditions, and 960 source pulses. Each condition had five calibration pulses and five later test pulses after a change in recorded stimulation contact order. Linear models used only completed earlier calibration sets in the same patient. Adding 20–100 ms log RMS to pre-pulse log RMS lowered 100–500 ms forecast error by 7.84% on average. All eight patient gains were positive. The descriptive 95% patient bootstrap interval was [6.16, 9.53]%. The exact two-sided patient sign-flip value was 0.0078125. An exact 500 ms shift of the windows gave a before-pulse control gain of 7.21%. The post-minus-control interval included zero. Thus the forecast gain does not establish a stimulation-specific effect. The tested question had no clear mean benefit in a prior ds004696 study. The present result has a different dataset and specified design changes. The package preserves six-window measurements, every forecast and fit, three figures, source code, source versions and hashes, native audit samples, and a streamed retrieval recipe. All 960 raw ranges passed source checks. Independent raw measurements, uncentered least-squares fits, timing checks, and a repeated analysis passed. This is a within-run, retrospective test of total recorded voltage. It does not test an online controller or a useful brain output.

stereoelectroencephalography · brain-computer interface · electrical stimulation · voltage forecast · root mean square · forecast error