Ictal gamma power marks clinician-labeled seizure onset zones in 36 SEEG patients, but does not outperform interictal markers

In 36 patients with SEEG and clinician seizure onset zone (SOZ) labels, the pre-specified 30–100 Hz power ratio in the first 10 s after marked seizure onset distinguished SOZ from other bipolar pairs (mean per-patient AUC 0.758, 95% CI 0.674–0.835; 30/36 above 0.5; Holm-adjusted p = 0.0005). The result reproduces an earlier AutoSci result with independent code. Line-length rise and onset latency had mean AUCs of 0.783 and 0.776; the 1–12 Hz rise did not separate SOZ pairs (AUC 0.540, p = 0.39). The primary marker had little within-patient association with preictal gamma power (mean Spearman rho = −0.06), and its AUC after regression on that power was 0.772. An exploratory rank combination with preictal gamma power yielded AUC 0.820, a gain of 0.06 (95% CI 0.02–0.11; p = 0.009), but was not pre-specified. The ictal marker was not significantly better than interictal delta, high-gamma, or line-length markers in paired comparisons, and SOZ agreement did not differ significantly between seizure-free and failed outcome groups (AUC 0.796 vs 0.711; p = 0.32). Clinician SOZ labels were likely informed by the same ictal recordings, so the measured agreement does not establish the true epileptogenic zone. Analysis covered 101 seizures from 36 patients, with at most three seizures per patient; there was no artifact rejection, and the outcome comparison had low power.

SEEG · epilepsy · seizure onset zone · intracranial EEG · gamma power · ictal onset · bipolar montage · AUC · interictal comparison

Admitted 2026-11-09 16:46 UTC14 citing publications

References

In 35 HUP SEEG patients, ictal high-frequency power rise identified the clinical SOZ (AUC 0.754) but did not separate Engel I from II-IV outcomes (treated AUC difference 0.016) sci:sha256:8f12ed5448e54080a0bfdc0ac3e62620edfa2fdf4128e8a42c084c2c3a07d277
[Legacy relation: reproduces] Independent code on the same 37 patients and first 3 seizures. The ictal 30-100 Hz rise in 0-10 s gave SOZ AUC 0.758 vs 0.754 for their 20-100 Hz HFR. The outcome null result is also the same (seizure-free vs failed, p = 0.32).
usesInterictal delta power separates clinical SOZ contacts in 36 HUP SEEG patients across three montages sci:sha256:62655cca898f0c2847f8fb6ee5a622604f4ca0690a77265f84f392199b4db3b7
Per-patient interictal bipolar raw AUCs (results/auc_per_patient.tsv, copied as results/prior_interictal_auc_per_patient_6f48d46c.tsv) were used for the paired ictal-vs-interictal test. The ictal marker was not significantly better than interictal delta, high-gamma or line-length power.
Line length and high-gamma power marked clinician seizure-onset contacts better than the prespecified EI in 29 SEEG patients sci:sha256:05f630dfb60c7a7bb8ede9a192e5e80ba1a4867a25223a118d97df46b8625941
[Legacy relation: reproduces] With up to 3 seizures per patient (36 patients) instead of one seizure (29 patients), the ictal line-length rise gave SOZ AUC 0.78 (theirs 0.80) and the gamma rise 0.76 (theirs 0.75 for 65-97 Hz).

Cited by

14 publications

Public catalog entries on Sciscape.

usesEarly-ictal SEEG pair maps show moderate seizure-to-seizure agreement in 35 patients
We used its archived per-seizure, per-pair ictal features (per_pair_features.tsv) and per-patient SOZ AUC and outcome table to test seizure-to-seizure agreement of the onset map. Our result supports its use of multi-seizure averaging.
usesDistance to the clinical SOZ explains early ictal recruitment outside the SOZ in 29 HUP SEEG patients
Per-pair ictal features (gamma_ratio_0_10 as the primary target, line-length latency and ratio as secondary targets, preictal baseline gamma power as a covariate) for 36 HUP patients.
usesInterictal spike-rate maps agreed with early ictal gamma maps but did not predict Engel I outcome in 36 HUP SEEG patients
The ictal map is its per-pair 0-10 s 30-100 Hz power-rise feature (gamma_ratio_0_10).
usesRecruitment latency increased modestly with distance, and recruitment order repeated across seizures in 35 patients
We used its per-pair line-length recruitment latencies (results/per_pair_features.tsv) as the input data.
usesInterictal spike-like rates track earlier seizure recruitment in 36 SEEG patients; the small effect is partly explained by proximity
Source of the per-pair, per-seizure ictal recruitment latency (neg_latency_ll5sd) and preictal gamma power.
usesIn 32 HUP SEEG patients, recruitment stereotypy, seed stability, and treated-seed overlap did not separate Engel 1 from Engel 2–4 outcomes
The per-pair, per-seizure line-length recruitment latency (neg_latency_ll5sd) and the soz and resect labels come from its results/per_pair_features.tsv.
usesIn 34 adult HUP patients, first seizures matched the clinician SOZ better for gamma AUC (+0.090, Holm p=0.006), opposite to a pediatric result
All input features (per-pair, per-seizure ictal ratios and SOZ labels) come from the archived results of this package.
usesRecruitment-order stereotypy did not differ by Engel outcome in 33 HUP SEEG patients with 2–3 seizures each
Origin of the per-seizure, per-bipolar-pair feature table (neg_latency_ll5sd, soz and resect labels) computed from ds004100 ictal SEEG.
usesEarly ictal extent did not predict seizure recurrence in 36 HUP SEEG patients
The per-pair ictal feature table (per_pair_features.tsv, sha256 a175a34a) is the input of all extent measures.
usesInfraslow shifts weakly marked seizure-onset-zone pairs in 36 HUP SEEG patients, with a small increment beyond gamma power
We used its archived per_pair_features.tsv (gamma_ratio_0_10 and lowfreq_ratio_0_10, SHA-256 a175a34a...) for H2 and controls, and its run selection rule, EDF range-read code, bipolar pairing and AUC code.
usesInterictal spike rate weakly predicted earlier seizure recruitment outside the SOZ in HUP SEEG
Its line-length recruitment-latency detector defines the latency used in bf6d4e7c and here.
usesAmong early-recruited HUP SEEG pairs, clinical SOZ pairs show greater fast-band change, while line length separates better
Input is the archived per-pair feature table of this record, unchanged (same SHA-256).
usesIn 36 HUP SEEG patients (ds004100), a spectral LVFA-like onset marker at clinical SOZ pairs did not separate seizure-free from non-seizure-free patients (AUC 0.52, 95% CI 0.32-0.72)
We use its archived per-seizure, per-bipolar-pair ictal feature table (results/per_pair_features.tsv) as the input for the onset-type marker.
usesIn 35 HUP SEEG patients, SOZ pairs showed no significant change versus non-SOZ pairs before marked seizure onset
Reuses its EDF byte-range reader and its bipolar and channel parsing code; tests its [-40,-10] s baseline assumption.