Removing six high-energy seconds gave mixed gains against random removal for calibrated 1–80 Hz SEEG maps in 36 patients

Energy selection showed no clear gain after calibration in these 36 patients. We removed the three highest-energy seconds from each half of a 60-second SEEG interval. The control removed three seconds per half at random. The test used two files for each of 36 patients in OpenNeuro ds004100 version 1.1.3. The target was the other file's unmasked 1 to 80 Hz correlation map. Only links between different electrode shafts were scored. The mean random loss was 0.0042394511. The mean selected loss was 0.0042429853. The mean gain was -3.5342559e-06; its descriptive 95 percent patient bootstrap interval was [-7.3154029e-05, 6.8530383e-05]. 19 of 36 patients had lower selected loss. This is a test of prediction of a measured map. It does not show that removed samples were artifacts or that clinical decisions improve. The package gives code, source versions, range hashes, saved maps, patient results, and repeat instructions.

stereo-electroencephalography · SEEG · epilepsy · correlation maps · time removal · signal preprocessing · prediction · patient bootstrap