32-second samples gave higher contact set overlap than 4-second samples against a separate minute for 80 to 150 Hz SEEG power with a common average reference in 19 interictal records

We tested how sample duration changes contact selection in public stereoelectroencephalography (SEEG) data. The source was OpenNeuro ds003876, version 1.0.2. Each of 19 participants supplied a two-minute interval from one interictal record. We used a common average reference and median epoch power from 80 to 150 Hz. We selected contacts in the top 10 percent by power, with the set size rounded up. A separate 60-second minute supplied the comparison set. At matched sample starts, mean set overlap was 75.3 percent for 4-second samples. It was 84.5 percent for 32-second samples. The mean increase was 9.2 percentage points. The 95 percent participant bootstrap interval was 7.4 to 11.0 points. All 19 participants had an increase. The exact two-sided sign test gave p = 0.00000381. For two full minutes, mean rank correlation was 0.975 and mean set overlap was 84.8 percent. Thus, high rank agreement did not ensure equal contact selection. Fresh source retrieval reproduced all 19 power arrays exactly. The package contains code, power arrays, source versions, hashes, results, figures, and repeat instructions. This result concerns selection within short interictal records. It does not establish clinical accuracy or a universal sample duration.

stereoelectroencephalography · SEEG · intracranial EEG · contact selection · high-frequency power · sample duration · set overlap · rank correlation · interictal records