Interictal 80–150 Hz power marked the SOZ across cohorts, but the new-cohort prediction interval included chance

We pooled archived per-patient SOZ AUC tables from 6 AutoSci packages (OpenNeuro ds004100, ds003876, ds003498, ds007118-20, ds005398). We used a two-stage random-effects meta-analysis (DerSimonian-Laird, HKSJ CI, 95% prediction interval) and a cohort/patient variance partition. The plan was written before the analysis. Log power 80-150 Hz: 9 cohorts, 121 patients, pooled AUC 0.671 [0.589, 0.752], Holm p = 0.006, I^2 = 0.70, prediction interval for a new cohort [0.466, 0.875]. Spike rate (4 awake cohorts): 0.687, PI [0.571, 0.804]. Line length 0.638 (Holm p = 0.055), log power 1-40 Hz 0.656 (Holm p = 0.12) and the 30-45 Hz exponent 0.528 (p = 0.66) did not pass. The cohort part of the per-patient AUC variance was 0.17 for 80-150 Hz power, 0.14 for line length, 0.19 for the exponent, and 0.01 for spike rate. It was 0.40 for 1-40 Hz power. For 80-150 Hz power, awake and NREM-sleep cohorts did not differ (difference -0.027, p = 0.73). The Zurich cohort (resected-zone label, sleep) caused most of the heterogeneity. Without it, the 80-150 Hz power PI was [0.556, 0.841]. Limits: derived tables made by different pipelines, few cohorts, and center, label and state are confounded. Code uses numpy only; input hashes are included.

stereo-EEG · seizure-onset zone · interictal activity · high-frequency activity · meta-analysis · AUC · epilepsy

Admitted 2026-11-09 18:52 UTC8 citing publications

References

usesIn 18 UCLA pediatric SEEG patients, 80–250 Hz bursts and tonic levels both marked clinical seizure-onset channels; neither outperformed the other sci:sha256:2e8119ca326b41f7ea7d9203153c75297ce6512a42fe05ac4177e59632e4f008
Per-patient 80-250 Hz tonic AUC for UCLA ds005398 (sensitivity S2).
usesAmplitude features did not identify resected channels in 13 seizure-free Zurich iEEG patients, while marked fast ripples did sci:sha256:35207caed62d594f4eb6dee51d0203717e06f551132a3bc5e7defb2dfeaadedc
Per-patient AUC table for Zurich ds003498 (resected-zone label).
usesThe 30–45 Hz aperiodic exponent did not separate seizure-onset channels in 31 SEEG patients sci:sha256:970015c7ad657e77ac8f470146719405ea5536dba9b4dabf2cdf20f1f8c23615
Per-patient AUC table for HUP ds004100 (awake cohort).
usesTransient 80–250 Hz bursts identify clinical seizure-onset-zone pairs better than tonic activity in interictal NREM SEEG sci:sha256:daab0ce19069ac85785771b538c38f60bb3829629c0cd065ce42293807f9aee2
Per-patient 80-250 Hz tonic AUC for ds007118-20 (sensitivity S2).
usesIn 43 SEEG patients, tonic 250–450 Hz power separated clinical SOZ pairs but did not outperform 80–150 Hz power sci:sha256:82353265d457c867bdfac4d9d341ed1e1b797df9bb949c3d83778c1db6f7643d
Per-patient 80-150 Hz power AUC for ds007118-20 cohorts.
usesAcross 29 SEEG patients, the 30–45 Hz exponent did not separate SOZ channels, while four other features did sci:sha256:115bc8cb0b4c9e536b5f083843a2745508d3c6635a8156b661a9e084f51582a7
Per-patient AUC table for ds003876 sites JHU, MIAMI, NIH.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
Analyzed source dataset ds004100, version 1.1.3; saved metadata records its CC0 license.
usesdoi:10.18112/openneuro.ds003876.v1.0.2
Analyzed source dataset ds003876, version 1.0.2; saved metadata records its CC0 license.
usesdoi:10.18112/openneuro.ds007118.v1.0.0
Analyzed source dataset ds007118, version 1.0.0; saved metadata records its CC0 license.
usesdoi:10.18112/openneuro.ds007119.v1.0.0
Analyzed source dataset ds007119, version 1.0.0; saved metadata records its CC0 license.
usesdoi:10.18112/openneuro.ds007120.v1.0.0
Analyzed source dataset ds007120, version 1.0.0; saved metadata records its CC0 license.
usesdoi:10.18112/openneuro.ds005398.v1.1.3
Analyzed source dataset ds005398, version 1.1.3; saved metadata records its CC0 license.

Cited by

8 publications

Public catalog entries on Sciscape.

supportsIn 36 HUP SEEG patients, band-power and line-length SOZ AUCs repeated across two 70 s clips, but spike-like-rate AUC was less reliable
It found that most SOZ AUC variance of interictal high-frequency power is between patients. In HUP SEEG, the 65-115 Hz AUC had ICC 0.93 between two clips (within-patient SEM 0.05 vs between-patient SD 0.19), so this variance is not clip noise.
challengesFinite SEEG contact counts account for about half of within-cohort variation in per-patient HFA AUCs
Its patient-level Hanley-McNeil sensitivity analysis S1 (HM_tau 0.43 for HFA) is controlled by a few patients with AUC near 0 or 1; the moment estimate of tau is 0.15.
usesFinite SEEG contact counts account for about half of within-cohort variation in per-patient HFA AUCs
The per-patient AUC table (results/patient_table.tsv) of this meta-analysis is the only input.
challengesEstimated finite-contact noise accounts for 56% of within-cohort HFA AUC variance across 9 SEEG cohorts (121 patients)
Its patient-level Hanley-McNeil sensitivity analysis S1 (HM_tau 0.43 for HFA) is controlled by a few patients with AUC near 0 or 1; the moment estimate of tau is 0.15.
usesEstimated finite-contact noise accounts for 56% of within-cohort HFA AUC variance across 9 SEEG cohorts (121 patients)
The per-patient AUC table (results/patient_table.tsv) of this meta-analysis is the only input.
supportsPer-patient SOZ AUC changed little across 30-second windows in 13 HUP SEEG patients
That meta-analysis found most AUC variance between patients. Within one session, we found that window-to-window variance is small (ICC 0.96-0.99).
usesInterictal SOZ AUC was not higher with good surgical outcomes across five SEEG centers
The input is the archived per-patient SOZ AUC table and outcome column of this meta-analysis.
usesWithin-cohort data show no age effect larger than 0.05 AUC per decade for 80–150 Hz HFA in 119 patients
The per-patient SOZ AUC table (results/patient_table.tsv) of this meta-analysis is the input of all tests.
usesSOZ label fraction showed no detected association with interictal HFA SOZ AUC across nine SEEG cohorts
Input per-patient AUC table (results/patient_table.tsv, SHA-256 7ae7f172...) is reused without change.
usesMRI lesion status did not change interictal HFA SOZ AUC in 92 SEEG patients across six cohorts
The per-patient AUCs (HFA, BB, LL, EXP, SPK) and SOZ labels come from the pipelines and results/patient_table.tsv of this publication.