Shared removal gave mixed 20–500 ms SEEG forecast effects and raised error on shafts with simulated gain 4 faults after five calibration pulses in eight people

We tested one shared pulse removal decision against separate decisions for each recording pair in public SEEG data from eight people with epilepsy. We used 85 stimulation sites and 879 recording pairs. The first five good pulses supplied calibration data. The final five supplied targets. The shared rule used only calibration data and gave each recording shaft equal weight. Both rules scaled the mean of four retained waveforms. For 20 to 500 ms forecasts, shared removal reduced error by a mean of 0.046 percent on recorded data and 0.050 percent with one gain 4 error at all recording pairs. Both descriptive intervals included zero. The shared rule removed the imposed pulse error in 99.20 percent of tests, compared with 97.14 percent for separate removal. When a gain 4 error affected only one shaft, shared removal raised affected-shaft error by 1.44 percent. Error rose in seven of eight people. The whole-panel score showed little change. This result shows why a calibration test must also score affected recordings separately. The package retains exact compact data, source pins, figures, decisions, scores, and checks. GPL analysis and extraction scripts remain in the local historical archive and are excluded. The imposed errors changed processed waveforms, not a device or stimulation current.

stereoelectroencephalography · calibration · electrical stimulation · forecasting · pulse removal · signal processing