Shared line-frequency estimation increased mean Brier error by 0.181 in 500-sample buffers from eight SEEG records with different pair frequencies and added-signal RMS 0.25× training RMS

We tested one line-frequency estimate shared by eight SEEG pairs against separate estimates in 500-sample buffers. The public data came from eight people in OpenNeuro ds004473 version 1.0.2. Added signals had a fundamental near 60 Hz and its second harmonic. Their combined expected RMS was one quarter of centered training RMS. With the same frequency across pairs, mean shared-minus-separate Brier error was -0.00080 (descriptive 95% person bootstrap interval -0.00246 to 0.00067; sign-flip p = 0.44531). The primary test did not show a clear benefit. A planned control assigned alternating fundamentals of 59.5 and 60.5 Hz across pairs. Shared mean Brier error was then 0.36545, compared with 0.18490 for separate estimates. The paired increase was 0.18055 (interval 0.13369 to 0.22953), with an increase in all eight records. This shows a limit of the shared-frequency assumption in this reader and added-signal model. The cohort and test trials had prior use. The package preserves exact selected voltage, source pins, all conditions, code, models, scores, figures, independent checks, and an exact full replay. It does not establish natural artifact frequencies, clinical benefit, or stimulation performance.

SEEG · line frequency · frequency estimation · signal processing · Brier error · brain-computer interfaces · synthetic interference