Minimum rank selection gave no detected precision change at a 10 percent count in 36 patients' two 120-second SEEG records under independent symmetric effects

We tested whether high contact ranks in both records improve seizure onset label precision. We used public OpenNeuro ds004100 version 1.1.3 and preserved window powers from an earlier AutoSci study. The matched set had 4,252 contacts from 36 patients, with two 120-second interictal SEEG records each. Both rules selected 10 percent of each patient's contacts, rounded up. The score compared power density at 2 to 4 Hz with density at 0.5 to 1 Hz. Mean patient precision was 16.21 percent for average ranks and 15.82 percent for the lower of two ranks. The paired change was -0.39 percentage points. Its 95 percent patient bootstrap interval was -1.78 to +0.98 points. The exact two-sided sign-flip test gave p = 0.59375 under independent patient effects with sign symmetry. These results do not establish equal precision. This is a new analysis of the same cohort, not an independent cohort replication. The package preserves code, inputs, contact choices, all bootstrap draws, source provenance, raw examples, figures, and reproduction instructions.

stereoelectroencephalography · seizure onset zone · contact selection · rank aggregation · precision · reproducibility