A 56-second SEEG threshold test on known contacts did not detect a mean balanced accuracy loss between recordings in 36 patients

A fixed threshold test used public SEEG data from 36 patients in OpenNeuro ds004100, version 1.1.3. Each patient had two recordings. The cohort had 4,252 matched contacts. Each signal sample contained 56 seconds. The fixed score was log10 of median 2 to 4 Hz density divided by median 0.5 to 1 Hz density. Mean balanced accuracy was 0.582 within recordings and 0.570 between recordings. The mean within minus between difference was 0.0121. Its patient bootstrap 95 percent interval was -0.0100 to 0.0339; the two-sided sign-flip p value was 0.297. The primary test did not establish equivalent accuracy. Apparent fitting accuracy was 0.699. A planned target median adjustment increased between-recording accuracy by 0.0195, with a descriptive interval of 0.0036 to 0.0357. Some reference and band sensitivity intervals excluded zero. The same known contacts and clinical labels entered fitting and testing. New patients, new contacts, and treatment outcomes were not tested. All 43 generated analysis files reproduced byte for byte. The package preserves code, source data provenance, decisions, random draws, figures, checks, and retrieval instructions.

stereoelectroencephalography · SEEG · epilepsy · threshold transfer · balanced accuracy · interictal EEG · signal repeatability · seizure onset zone