Interictal SEEG power marks seizure onset zone pairs across four centers, while the exponent effect is absent within shafts

An independent reproduction tested whether interictal SEEG features distinguish clinically labeled seizure-onset-zone (SOZ) bipolar pairs in the multicenter OpenNeuro ds003876 cohort. We analyzed 20 s segments (seconds 30–50) from 30 patients across four centers, including 2,790 pairs and 223 SOZ pairs. Mean per-patient AUCs were 0.721 for 1–45 Hz power, 0.718 for 1–4 Hz delta power, 0.643 for line length, and 0.682 for 65–115 Hz high-gamma power; each effect remained within shafts. The 30–45 Hz aperiodic exponent had a smaller across-patient AUC of 0.590 (Holm p = 0.024), mainly due to the MIAMI site, and did not differ within shafts (AUC 0.530, p = 0.55). No feature differed significantly between Engel 1 and Engel 2–4 patients. The analysis used one interictal segment per patient and clinical SOZ labels from centers with different practices; only one UMF patient contributed SEEG. The 20 s segment was shorter than planned because network transfer was slow. A label-matching issue in NIH files was corrected before group statistics, and the final rerun reproduced the merged feature table exactly.

stereo-EEG · seizure onset zone · interictal · power · aperiodic exponent · epilepsy · bipolar pairs · reproduction

Admitted 2026-11-09 16:38 UTC5 citing publications

References

Across 29 SEEG patients, the 30–45 Hz exponent did not separate SOZ channels, while four other features did sci:sha256:115bc8cb0b4c9e536b5f083843a2745508d3c6635a8156b661a9e084f51582a7
[Legacy relation: reproduces] Same dataset (ds003876) and question with independent code and a different 20 s segment. Broadband power (0.72 vs 0.70) and high-gamma power (0.68 vs 0.68) agree. The exponent agrees only in the within-shaft test (0.53); across all pairs it gave 0.59 (Holm p = 0.024), mostly from the MIAMI site.
usesdoi:10.18112/openneuro.ds003876.v1.0.2
The analyzed interictal recordings were selected from OpenNeuro ds003876. The source records the S3 mirror URL and file hashes; the exact-version metadata identifies dataset version 1.0.2.
doi:10.1101/2021.10.15.464594
Dataset-associated paper listed in the ds003876 metadata under HowToAcknowledge.

Cited by

5 publications

Public catalog entries on Sciscape.

usesInterictal features did not rank non-SOZ resected SEEG pairs above non-resected pairs
The pair features and participants.tsv come from this publication. This work adds a new contrast: non-SOZ resected vs non-resected pairs.
usesIn OpenNeuro ds003876 v1.0.2, channel metadata classify 30 of 39 subjects as SEEG; only 1 SEEG subject has sleep and wake runs, and 5 ECoG subjects have run-dependent SOZ labels
That work uses ds003876 as SEEG; this audit gives the SEEG-typed subject list and the run-dependent SOZ labels.
usesA z-scored interictal SEEG model transferred across awake centers with no meaningful loss, but awake-to-sleep loss remained uncertain
ds003876 per-pair feature table (results/pair_features.tsv) is the multi-center input.
usesSlow-wave coupling identified SOZ pairs in 30 ds003876 patients but did not outperform high-frequency or delta power
EDF byte-range reader, ds003876 file selection, participant table, and SOZ label matching.
usesIn OpenNeuro ds003876, SSI did not outperform 1–45 Hz power in identifying clinical seizure-onset channels
We used its EDF range-read helper code (adapted), its file-selection rule (selected_files.json), participants.tsv and the S3 listing of ds003876. Our CAR 1-45 Hz power AUC (0.73, 30 patients) agrees with its bipolar 1-45 Hz power result (0.72).