Neural Noise- A normative iEEG reference improves detection of epileptic abnormalities across wakefulness and sleep.1
- State-dependent normative data support clinical neurophysiology and iEEG analysis methods (medRxiv preprint).1 1
Weekly enrichment (2026-07-20)
- The medRxiv preprint (posted September 2025) built a normative intracranial-EEG atlas spanning wakefulness plus N2, N3, and REM sleep, extending prior wake-only atlases to state-dependent references.2
- It analyzed interictal iEEG from 30 drug-resistant epilepsy patients against a normative reference drawn from 106 additional patients (overall DRE cohort of 108; 29 primarily ECoG grids, 76 primarily SEEG depth electrodes).2
- Spectral power and coherence were computed across six canonical frequency bands and converted to region-matched z-scores, with the validated SleepSEEG algorithm staging a 12-hour night-time recording per patient.2
- A leave-one-subject-out cross-validated random forest scored each channel’s abnormality; channel-level normal-versus-abnormal classification AUC was highest in REM sleep (0.86, 90% CI 0.81-0.90) versus N2 (0.72), N3 (0.70), and wake (0.69).2
- Patients seizure-free at two years (Engel 1) showed higher separation of within- versus outside-resection abnormality (AUPRC) than poor-outcome patients for N2-sleep models (Mann-Whitney, p = 0.003).2
- Combining wakefulness and NREM (N2) features best separated good from poor surgical-outcome subgroups (Cohen’s d = 1.62).2
- Combined wake/sleep abnormalities plus interictal spike rates explained 57-58% of the variance (R² = 0.57-0.58) in presurgical neuropsychological test performance.2
- Permutation-based feature importance ranked broadband and gamma coherence highest and delta/theta bandpower lowest, and about 5 minutes of iEEG per state was enough for stable abnormality estimates.2
- All iEEG data and analysis code were released publicly (Pennsieve and GitHub), supporting multicenter replication and prospective testing.2
- Clinical stakes: epilepsy affects over 50 million people, 20-30% are drug-resistant, and postoperative seizure freedom is only 34-74% (median 62.4%), so interictal, sleep-inclusive localization could shorten epilepsy-monitoring-unit stays and reduce provoked seizures.2 3
Footnotes
-
https://news.google.com/rss/articles/CBMie0FVX3lxTE1TUk1zWWhNM1dEbjEyV1N5UFhmcGxiTXhyRWc1dnN1N2pvUURXYUlmdWEyN2VwWjhPbU5ZX1p5V0lUcVM1QmZpbHRZdkljNjNfc1lGNE1iWXFjMEhFVmhHZWJWdHJNNXB4WWdCbzFELUQxV0h5TWEyS3NBMA?oc=5 ↩ ↩2 ↩3
-
https://www.medrxiv.org/content/10.1101/2025.09.10.25335533v1.full-text ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10