A 60 Hz trend fit reduced 100 ms gamma power error in known SEEG test signals with a 60.5 Hz disturbance at peak 10 RMS

We tested two time terms in a local 60 Hz fit for short gamma power windows. We made known 70 to less than 110 Hz signals from 72 public SEEG records in 36 epilepsy patients. The main test used 100 ms windows and an added 60.5 Hz sine. Its peak voltage was ten times the known signal RMS. Errors were averaged over four fixed phases. The time terms reduced mean normalized power MSE from 0.027987 to 0.003394. The reduction was 87.87%, with 95% patient bootstrap limits of 87.03% to 88.56%. All 36 patient signals had less error. Without an added disturbance, the same terms increased known-signal power error by 2.22 times. Both fits could raise measured gamma power. The simpler fit was better at exactly 60 Hz and with a small disturbance. The package supplies a controlled benchmark, a local fit implementation, and three figures. It preserves native digital data, pinned source versions, all scores, code, a fixed local plan, and reproduction instructions. All 72 data archives rebuilt byte for byte from fresh public source ranges. A full replay reproduced all scientific outputs. The known signals use an offline construction. The result does not measure intention decoding, clean neural signal recovery, or stimulation benefit.

SEEG · gamma power · line interference · signal processing · local regression · Hann window · known-signal benchmark