The STE spike selection AUC change interval included zero within shafts in the first hour of 17 SEEG records under independent, exchangeable patient sampling

We tested whether released spike scores improve local seizure onset channel ranking above counts selected by artifact scores. We used public sleep SEEG from OpenNeuro ds005398 version 1.1.3. The fixed primary test used STE events in the first hour and a spike threshold of 0.5. It included 474 SOZ versus other channel pairs on 38 shafts in 17 patients. Mean patient AUC was 0.645847 for HFO counts and 0.647391 for spike HFO counts. The paired mean change was 0.001544. Its nominal 95 percent patient bootstrap interval was -0.035453 to 0.039583 under independent, exchangeable patient sampling. The interval included zero. The test did not establish a gain or equivalence. All 20 fixed comparisons are reported. MNI results show that score selection can affect detector counts and ranks differently. The package preserves public input bytes, versions, hashes, new code, count and pair tables, patient results, a figure and repeat instructions. Fresh public checks matched 94 input objects or EDF headers. A separate parser checked 38,550 counts. Repeated analysis matched seven numerical files exactly. This reused cohort does not establish external classifier performance or clinical benefit.

high-frequency oscillations · stereoelectroencephalography · seizure onset zone · channel ranking · within-shaft comparison · patient bootstrap