Nominal 95 percent bootstrap bands had higher mean contact failure across SEEG records than within records in 35 patients with 30 second centered high gamma maps and common average reference

We tested nominal 95 percent prediction bands for public SEEG power maps in 35 patients. We used two interictal records per patient from OpenNeuro ds004100 version 1.1.3. Each training and test map used 30 seconds. The primary maps used common average reference and centered 80 to 150 Hz power. Training windows alone set the bands. The mean contact failure was 30.43 percent in the adjacent half and 57.67 percent in the other record. The mean increase was 27.25 percentage points. Its 95 percent patient bootstrap interval was 21.79 to 32.85 points, under independent patient sampling. Thirty-four patients had a positive difference. All fixed method checks had positive group differences. A stationary independent window model also failed the nominal target. It gave 11.40 percent distant contact failure. Thus, measured failures do not establish a change in brain activity. This is a new endpoint in an existing selected clinical cohort. The package preserves exact inputs, source versions, code, bands, contact decisions, resampling draws, model checks, and repeat instructions. Unknown source record separation and artifacts limit interpretation. All 70 raw source prefixes passed independent power checks. Complete repeats from saved powers and fresh raw data matched the archived analysis.

prediction bands · bootstrap · stereoelectroencephalography · SEEG · high gamma power · coverage · signal processing · reproducibility