Relative beta power increased threshold error in three people and reduced error in four after five chronic SEEG calibration sessions

We compared relative beta power with absolute beta power in public chronic hippocampal SEEG from seven people. Each person had separate fixed thresholds set from the first five selected sessions. Each threshold had a 10 percent rate target. We tested 187 later sessions with twenty two-second epochs per session. Mean absolute rate error was 12.24 percentage points for absolute power and 10.76 points for relative power. Each person had equal weight. The change was -1.48 points. The descriptive 95 percent person bootstrap interval was -5.33 to +2.17 points. Relative power gave lower error in four people and higher error in three. All planned method checks are reported. The second-channel check gave lower error in six people, but this was a secondary test without a multiple-comparison adjustment. Exact cancellation of constant voltage gain did not guarantee lower observed error in each person. The package contains all 222 selected EDF files, source metadata, a local plan, code, results, and three figures. All 222 EDF files and 666 sidecars passed fresh source checks. A separate reader agreed on all 8.14 million voltage samples. A separate run reproduced seven result files exactly. These event-triggered clips lack confirmed seizure and intention labels. Rate error is not false alarm rate, decoding accuracy, stimulation benefit, or clinical benefit.

relative beta power · threshold transfer · stereoelectroencephalography · brain-computer interface · chronic intracranial recording · signal processing