A scale from 24 early pulses reduced later centered SEEG power more than full subtraction at 3 to 15 ms in 20 stimulation trains

We tested a scale for pulse template subtraction in public SEEG data from OpenNeuro ds005169 version 1.0.0. We used one stimulation train from each of 20 people and all 1,671 eligible stored bipolar channels. The first 24 pulses set a signed class template. Agreement between two early templates set its scale. The next 24 pulses tested the fixed correction. At 3 to 15 ms after detected edges, scaling gave 2.28 microvolts squared more mean power reduction than full subtraction. The 95 percent person bootstrap interval was 0.94 to 4.44. The difference was positive in 18 people. Full subtraction added power in nine people; scaling added power in three. At matched baseline, the advantage was larger: 3.16 microvolts squared. The stimulation minus baseline difference was -0.88, with an interval from -2.23 to 0.81. A fixed pulse grid gave a similar result. This exploratory study reused selected epochs from a prior publication. It supports a method to limit power added by uncertain templates. It does not establish removal of artifact or improved neural decoding. The package includes three figures, exact voltage epochs, code, fitted templates, results, source hashes, and reproduction instructions. Fresh source extraction and an isolated Python environment reproduced the saved numeric files.

stereoelectroencephalography · SEEG · template subtraction · stimulation artifact · signal processing · intracranial EEG · pulse calibration · brain-computer interface