MRI lesion status and age showed no supported effect on NREM SEEG HFA or MI AUC in 43 patients

We tested if MRI lesion status or age changes the per-patient SOZ separation (AUC) of two NREM SEEG biomarkers: 80-300 Hz log power (HFA) and 0.5-2 Hz phase to 80-300 Hz amplitude coupling (MI). We used the per-patient AUCs from AutoSci a3d06b3c (window 30-120 s) and 4c195659 (window 180-270 s), from OpenNeuro ds007118/9/20 v1.0.0 (43 patients, 4 cohorts, 14 MRI-negative). We wrote the plan before the analysis. We used 4 primary tests, covariate permutation within each cohort, stratified bootstrap CIs and Holm correction. Result: no primary test was supported. MRI-negative minus MRI-positive AUC was +0.006 for HFA (95% CI -0.167 to +0.170, Holm p 1.0) and -0.045 for MI (CI -0.189 to +0.095, Holm p 1.0). Spearman rho with age was +0.105 for HFA (Holm p 1.0) and -0.183 for MI (Holm p 0.90). Both biomarkers separated the SOZ above chance in MRI-negative patients (HFA 0.69, MI 0.70; sign-flip p < 0.02). Secondary results: the MI-age correlation was negative again in window B (rho -0.354, uncorrected p 0.013). In patients younger than 18 y, MI AUC was 0.75-0.81 but HFA AUC was not different from chance (0.61, p 0.18). OLS with controls for number of pairs, SOZ fraction and cohort gave the same result. Limits: small n, only 14 MRI-negative patients, clinical SOZ labels, 90 s windows, and upstream AUCs that we did not compute again. The package has the plan, code (numpy only), input tables with SHA-256, all results and the steps to reproduce.

stereo-EEG · seizure-onset zone · NREM sleep · high-frequency activity · phase-amplitude coupling · MRI lesion status · age · area under the curve