Median SEEG power from 10 seconds did not meet a 5 percentage point agreement rule against mean power from 60 seconds in 34 patients

We tested whether less SEEG data can preserve agreement of lists of pairs with the highest power. We used 3,325 fixed bipolar pairs from 34 patients in OpenNeuro ds004100 version 1.1.3. Power covered 80 to 150 Hz, with 118 to 122 Hz removed. We compared the median of ten one-second windows with the mean of sixty windows. Each method selected the highest 10 percent of pairs in each of two source runs. Mean list agreement was 74.45 percent for the short median and 77.67 percent for the long mean. The paired difference was -3.21 percentage points. Its 95 percent patient bootstrap interval was -6.66 to +0.22 points. The lower bound was below the fixed -5 point limit. Thus the short method did not meet the local analysis rule. Against the median of 60 seconds, agreement was 8.35 points lower. Results changed with list size and frequency band. These checks are descriptive. The study uses a selected epilepsy cohort and does not show clinical benefit. All 72 pinned source prefixes were retrieved. All 68 power files and 36 quality records matched the parent archive. The package preserves the local plan, code, inputs, patient results, bootstrap indices, figures, and repeat instructions.

stereo-EEG · SEEG · high-frequency power · recording duration · pair-list agreement · signal processing · epilepsy