Median and maximum monitors had mean AUC 0.763 and 0.774 for the first 20 seizure seconds in 30 SEEG records after onset shafts were excluded

We tested whether a median across electrode shafts gives better seizure separation than the largest shaft score. We used 30 public SEEG records from OpenNeuro ds004100 version 1.1.3. We excluded every shaft with an onset label. We fit feature scales and alert thresholds in separate 30-second intervals. We tested 50 later pre-seizure seconds and the first 20 seizure seconds in each patient. Mean patient AUC was 0.763 for the median and 0.774 for the maximum. The paired mean difference was -0.011. Its descriptive 95 percent patient bootstrap interval was -0.044 to 0.021. The sign test gave p = 1.0. The median did not show better separation. An exact bound proved that 1,719 of 2,100 median decisions stay fixed after any one nonnegative shaft score replacement, with the threshold fixed. The median alert rate was 9.3 percent before onset and 48.8 percent during seizures. This is an exploratory extension of a cited cohort study. It does not test interface commands or stimulation. The package preserves powers, small raw subsets, fits, scores, code, figures, hashes, and source checks.

stereoelectroencephalography · seizure monitoring · signal processing · brain-computer interface · robustness · AUC · electrode shafts