Rationale:
Accurate localization of epileptogenic sources is essential for pre-surgical planning in focal drug-resistant epilepsy (fDRE) [1]. Electrophysiological source imaging (ESI) using scalp EEG offers a promising approach, with biomarkers like interictal spikes, high-frequency oscillations (HFOs) [2], and seizures [3] each contributing to the mapping of epileptogenic zone (EZ). However, the quantitative differences in source imaging performance among these biomarkers remain insufficiently explored, and current methods rely on separate processing pipelines which limit clinical integration. In this study, we developed a unified spatial-temporal-spectral source imaging (STSI) framework that jointly considers spatial, temporal, and spectral features of EEG signals. We applied STSI to assess and compare the source imaging performance of spikes, HFOs, and seizures in a cohort of fDRE patients.
Methods:
The STSI framework first extracts components representing the spatial, temporal, and spectral features of EEG biomarkers, then estimates the underlying sources by solving an optimization problem. We applied this framework to 2,081 epileptic events, including seizures, HFO-spike co-occurring events (pHFOs and pSpikes) [2], general spikes (aSpikes), and general HFOs (aHFOs), from a cohort of 42 fDRE patients (30 seizure-free, ILAE I–II; 12 non-seizure-free, ILAE III–V). Source localization accuracy was quantitatively evaluated against clinical references, including surgical resection areas and/or intracranial EEG-defined seizure onset zones.
Results:
ESI analysis results are presented in Fig. 1, including localization error (LE), spatial dispersion (SD), precision (Prec), sensitivity (Sen), specificity (Spec), and F1 score (F1). The STSI framework enabled quantitative comparisons across key EEG biomarkers, yielding average localization errors of 7.16 mm for seizures, 10.23 mm for HFOs overlapping with spikes (pHFOs), 12.22 mm for HFO-riding spikes (pSpikes), 23.16 mm for general spikes (aSpikes), and 37.71 mm for general HFOs (aHFOs) when compared with the surgical resection area.
Conclusions:
Our results indicate that seizure and pHFO are among the most accurate biomarkers, followed by pSpikes, aSpikes and aHFOs. These results highlight the merits of pHFOs as interictal biomarkers for EZ mapping and emphasize the clinical value of a unified, quantitative source imaging approach. The STSI framework represents a significant advancement toward standardizing ESI methodologies and holds promise for improving surgical outcomes in fDRE patients.
References:
1. He B, Sohrabpour A, Brown E, and Liu Z, 2018. Electrophysiological Source Imaging: A Noninvasive Window to Brain Dynamics. Ann Review of Biomedical Engineering, 20, 171-196.
2. Cai Z, Sohrabpour A, Jiang H, Ye S, Joseph B, Brinkmann BH, Worrell GA, and He B, 2021. Noninvasive high-frequency oscillations riding spikes delineates epileptogenic sources. PNAS, 118(17), p.e2011130118.
3. Ye S, Yang L, Lu Y, Kucewicz MT, Brinkmann B, Nelson C, Sohrabpour A, Worrell G, He B, 2021. Ictal Source Imaging Contributes to Seizure Onset Zone Localization in Focal Epilepsy Patients. Neurology, 96(3), E366-E375.
Funding:
NIH NS127849, NS096761