Fast power did not exceed baseline during 29 of 446 first sustained 6 dB ratio rises in 26 SEEG patients

We tested band power changes at the first sustained spectral ratio rise near seizure onset. Public SEEG data came from OpenNeuro ds004100, version 1.1.3. We used two seizures from each of 26 patients and 424 local bipolar pairs in 56 contact name groups. A detection required a 6 dB ratio rise in two consecutive one-second blocks within 20 seconds after source onset. Fast power was 20–50 Hz. Slow power was 2–12 Hz. There were 446 first detections. Fast power did not exceed its own pre-onset median in both blocks of 29 detections. Slow power was below its own median in both blocks of these detections. This occurred in nine patients. The mean patient fraction was 7.57%. The 95% patient bootstrap interval was 2.80% to 13.44%. This conditional estimate used the 25 patients with a detection. An added fast power rise requirement gave mean timing AUC 0.713, compared with 0.724 for the ratio rule alone. The paired change was −0.0112, with a 95% interval of −0.0322 to 0.0043. This did not show an AUC gain or establish equal performance. The result concerns measured signals and source clinical labels in a selected cohort. It does not identify a causal tissue change. We reused earlier AutoSci features and independently checked all new event calculations. All 52 fresh public source ranges matched their hashes. Rebuilt features matched all 658,080 saved power values exactly. The package preserves code, source records, features, results, figures, and repeat instructions.

stereoelectroencephalography · epilepsy · seizure onset zone · spectral power · power ratio · event detection · intracranial EEG · signal processing