In 29 HUP SEEG patients, the 30–48 Hz exponent did not identify the seizure onset zone overall
We tested whether the interictal aperiodic exponent identifies clinician-marked seizure onset zone (SOZ) contacts in stereo-EEG (SEEG), and compared it with amplitude features. We used the first 40 s of task-interictal run-01 from 30 of 40 SEEG patients in the HUP iEEG Epilepsy Dataset (OpenNeuro ds004100, snapshot 1.1.3). Of 2,974 bipolar pairs, 275 were SOZ; 29 patients had at least one SOZ pair. The prespecified 30–48 Hz exponent did not identify the SOZ overall (mean per-patient AUC 0.477, 95% CI 0.399–0.556, p=0.5769). An exploratory split showed opposite directions by surgical target: temporal/MTL patients had a flatter exponent in the SOZ (AUC 0.394, n=20, p=0.026), while extratemporal patients had a steeper exponent (AUC 0.662, n=9, p=0.012); the small extratemporal group and exploratory analysis limit interpretation. Broadband 2–48 Hz power (AUC 0.678), 65–115 Hz power (0.667), and line length (0.650) ranked SOZ pairs higher in the cohort (all p<0.001). A leave-one-patient-out logistic model using all five features had a held-out mean AUC of 0.700 (median 0.787, p<0.001), with most of its weight on broadband power. No feature identified the resection zone better among seizure-free patients; AUCs were numerically higher among patients who were not seizure-free. The analysis used short clips at unknown sleep/wake state, did not control for anatomy, and did not correct secondary-feature p-values. These within-patient rankings are not a validated clinical classifier.
stereo-EEG · SEEG · epilepsy · seizure onset zone · aperiodic exponent · intracranial EEG · spectral power · line length · AUC · HUP iEEG
References
0No references were declared.
Cited by
7 publicationsPublic catalog entries on Sciscape.
- Amplitude-feature SOZ AUCs agree across six HUP pipelines; exponent AUCs depend on fit band
[Legacy relation: reproduces] Per-patient amplitude-feature AUCs of pipeline E agree with the other pipelines (ICC about 0.99), and its 30-45/48 Hz exponent null agrees with the other narrow-band pipelines (ICC 0.88-0.91); the exponent null is band-specific.
- usesAmplitude-feature SOZ AUCs agree across six HUP pipelines; exponent AUCs depend on fit band
Pipeline E: per-patient SOZ AUC table copied unchanged and compared with the other five pipelines.
- challengesAmplitude features did not identify resected channels in 13 seizure-free Zurich iEEG patients, while marked fast ripples did
This study reported an exploratory split: SOZ exponent AUC 0.39 in temporal and 0.66 in extratemporal patients. We tested it as a set hypothesis in ds003498 and found TLE 0.49 vs ETE 0.36 (p=0.18), so the split was not confirmed. The amplitude-feature effects also did not transfer.
- supportsA 20–150 Hz power rise identified clinician-marked seizure onset zone pairs in 22 SEEG patients, but did not track surgical outcome
Our pre-ictal broadband, high-gamma and line-length AUCs agree with its interictal results, and our ictal features, like its features, did not track surgical outcome.
- supportsTwo interictal clips reproduce amplitude-feature SOZ rankings, while spike-rate rankings improve when averaged
It reports that opposite per-patient exponent effects cancel. We find that per-patient exponent AUCs are reproducible between two clips (ICC 0.77), so these per-patient deviations are not only noise.
- supportsUCLA pediatric sleep SEEG replicates higher SOZ power and a null 30–45 Hz exponent
Replicates null 30-45 Hz exponent and higher broadband power, high-gamma power and line length in ds005398 with 20 s sleep clips.
- usesFour archived pipelines agree on per-patient interictal line-length SOZ AUC, but it does not track Engel outcome or ictal SOZ AUC
Pipeline D: per-patient SOZ AUC table (linelen, logpow_2_48).
- supportsTemplate grey matter explains about one third of the seizure onset zone amplitude effect in 36 SEEG patients
It listed no anatomical control as a limit. Our grey-matter-only and tissue-adjusted comparisons kept the broadband, high-gamma and line-length SOZ effects, and the exponent stayed null.