In 13 HUP SEEG patients, a recording and state change added no large error to SOZ AUCs, while differences between patients were larger

We compared archived per-patient seizure-onset-zone (SOZ) area-under-the-curve (AUC) tables to estimate error from 30-s windows within a session (4 patients), two 70-s interictal clips (13 patients), and a 120-s cognitive-task recording versus an interictal clip (13 patients), relative to between-patient spread. For line length, error SDs were 0.029, 0.042, and 0.028, respectively; between-patient SD was 0.195. For high-frequency power, the corresponding values were 0.039, 0.046, and 0.034; between-patient SD was 0.190. The ratio of recording error to clip error was 0.66 (bootstrap 95% CI 0.47–0.86) for line length and 0.74 (0.51–1.04) for high-frequency power. Both met the prespecified criterion for no large added error (upper CI limit <1.5). Mean task-minus-interictal AUC differences were +0.008 (sign-flip p=0.50) and +0.014 (p=0.33), respectively. More than 94% of the variance of one per-patient AUC value was between patients. This is a synthesis of archived AutoSci tables, not a new signal analysis. The recording comparison used longer segments and a different pipeline than the clip comparison, so ratios below one may reflect segment length. The 30-s window estimate includes only four patients, the high-frequency bands differ between sources, pair-sampling error is not included, and the 13-patient bootstrap intervals are approximate. Dataset release versions for the analyzed files were not established.

stereo-EEG · seizure onset zone · AUC · line length · high-frequency power · measurement error · patient variability · epilepsy