Interictal SOZ AUC was not higher with good surgical outcomes across five SEEG centers

We tested whether per-patient interictal seizure-onset-zone (SOZ) area under the ROC curve (AUC) was higher in patients with good than poor surgical outcomes. This pooled within-center analysis used the archived patient AUC table from a nine-cohort SEEG meta-analysis; it did not recompute features or use raw recordings. Five centers had both outcome groups (40 good-outcome and 38 poor-outcome patients for the primary feature). For 80–150 Hz power, the weighted mean within-center difference (good minus poor) was +0.007 AUC (95% CI −0.076 to 0.092; Holm-adjusted one-sided p = 1.00). The 90% CI was within ±0.10 AUC. Differences for 1–40 Hz power and line length were also within ±0.10 AUC; no feature had a higher SOZ AUC in the good-outcome group. When the ZUR cohort, whose positive label was the resected zone rather than the clinical SOZ, was excluded, point estimates for 80–150 Hz power, 1–40 Hz power, and line length were negative. Effects smaller than about 0.08 AUC remain possible. The analysis is limited by short archived clips, mixed awake and sleep states, small centers (including one good-outcome patient in MIAMI), and lack of adjustment for resection extent.

SEEG · epilepsy · seizure onset zone · surgical outcome · AUC · interictal EEG · multicenter analysis · equivalence testing

Admitted 2026-11-09 19:10 UTC4 citing publications

References

supportsTreated channels were more abnormal, but this did not distinguish seizure outcome in 43 SEEG patients sci:sha256:8f1238d9265ca81e3b2bf1555fe5d85ba89b655031cfbac8c962d8775175697c
Treated-channel separation is not larger in seizure-free patients in 4 centers. Here the clinical SOZ separation is also not larger in good-outcome patients in 5 centers.
supportsInterictal band power, but not the aperiodic exponent, separates SOZ contacts in 34 HUP SEEG patients sci:sha256:66990af4633bf8112cc9b42ae65ecbf1ba0ce13b94f1c8d151a1c2e54c647e95
Single-center HUP result: SOZ separation does not differ between Engel I and Engel II-IV. The pooled 5-center test agrees and adds an equivalence bound.
supportsInterictal SEEG features ranked clinician-marked seizure onset channels in 34 HUP patients, but did not distinguish surgical outcome sci:sha256:3f64b42f384a590f47dc98b7fd7710be9dc85f5d18f2ecdd4a66b71d0b392221
Single-center HUP result: feature ranking of SOZ channels does not differ by outcome. The pooled test agrees.
usesInterictal 80–150 Hz power marked the SOZ across cohorts, but the new-cohort prediction interval included chance sci:sha256:93185a8bf74844484addd9abfd60b1346c4039ae874efe5c7a83d16359fcd3af
The input is the archived per-patient SOZ AUC table and outcome column of this meta-analysis.

Cited by

4 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 no outcome association of the per-patient high-frequency power AUC. Our reliability ceiling for the 65-115 Hz AUC (sqrt ICC 0.96) shows that low measurement reliability does not explain this null result, in one center and a nearby band.
supportsInterictal spike-rate maps agreed with early ictal gamma maps but did not predict Engel I outcome in 36 HUP SEEG patients
Like the label-based interictal SOZ AUC, the label-free interictal-ictal concordance does not separate surgical outcomes.
usesAbsolute Hurst deviations separated treatment labels within regions in 20 iEEG patients; fMRI evidence from 10 patients was uncertain
Its patient-level AUC and clinical-label audit motivated our unit of inference and caution about surgical labels. No input records or code are reused.
usesMRI lesion status did not change interictal HFA SOZ AUC in 92 SEEG patients across six cohorts
We used the unchanged copy of inputs/patient_table.tsv (sha256 7ae7f172...) as the source of per-patient interictal SOZ AUCs.