Interictal SEEG Lempel–Ziv complexity was lower in the SOZ but added no SOZ information beyond line length and variance

In 36 patients from the HUP intracranial EEG cohort (OpenNeuro ds004100), we assessed whether Lempel–Ziv complexity (LZC) in short interictal bipolar SEEG segments distinguished clinically labeled seizure-onset-zone (SOZ) pairs and added information beyond line length, variance, or the power spectrum. Across 3,790 bipolar pairs, mean per-patient LZC AUC was 0.371 (two-sided sign-flip p=0.0009; Holm-adjusted p=0.0027), with lower LZC in SOZ pairs in 75% of patients. However, LZC residualized on log line length and log variance had AUC 0.516 (p=0.54), and excess LZC over phase-randomized surrogates had AUC 0.457 (Holm-adjusted p=0.37). Surrogate LZC alone had AUC 0.391. On SOZ-bearing shafts, LZC did not distinguish SOZ from adjacent non-SOZ pairs (AUC 0.471), whereas line length did (AUC 0.714, p<1e-5). Under the preregistered interpretation rule, LZC re-expressed spectral and amplitude information and showed no additional SOZ information in this cohort. An uncorrected excess-LZC effect in 15 patients with six full windows (AUC 0.375, p=0.006) is hypothesis-generating; segment length and patient selection may explain it. The single short segment per patient, variable segment duration, one surrogate per window, median-only binarization, clinical labels, and reuse of the prior cohort limit interpretation; this is not an independent replication.

Lempel–Ziv complexity · stereo EEG · seizure onset zone · epilepsy · signal complexity · phase-randomized surrogate · intracranial EEG