Authors :
Presenting Author: Garnett Smith, MD – University of Michigan
Stephen Gliske, PhD – Neurosurgery – University of Nebraska Medical Center; William Stacey, MD, PhD – Professor, Neurology, University of Michigan
Rationale:
Quantitative biomarkers have the potential to improve outcomes in Epilepsy Surgery, especially in cases where there is no clear focal lesion on imaging. Two of these are Epileptogenicity Index (EI) and Slow Polarizing Shift index (SPS). Since EI relies on a robust increase in fast frequency ( > 12 Hz) power at the onset of a seizure, it is most helpful when the dominant frequency at seizure onset is above 12 Hz. In contrast, SPS, which is based upon quantification of DC shift at seizure onset, was designed to capture information in slower frequency bands. We compared the performance of EI and SPS in two groups of seizures: those with low amplitude fast activity at onset and those without.
Methods:
We analyzed 212 seizures from 22 patients. Seizure onset patterns were classified by separating seizures into eight patterns according to the presence or absence of low voltage fast activity (LFA), DC shift, or rhythmic spiking in various frequency bands. We next calculated Receiver Operating Characteristic (ROC) curves for these biomarkers to compare their performance.
Results:
The AUC was greater for fast-onset seizures for both biomarkers. For EI, AUC was 0.73 for fast onset seizures and 0.60 for slow-onset seizures (p = 0.0001). For SPS AUC was 0.66 for fast-onset seizures and 0.57 for slow-onset seizures (p = 0.0062). Overall, the highest AUC was obtained for EI in seizures with fast activity at onset. For fast onset seizures AUC for EI was significantly higher than AUC for SPS (0.04), but for slow-onset seizures AUCs for EI and SPS were not significantly different. The AUC for EI was significantly higher for fast-onset seizures (p = 0.0001), but AUC was not significantly different for SPS in fast versus slow seizures (p = 0.31).
Conclusions: Epileptogenicity Index outperformed Slow Polarizing Shift Index even in seizures that were classified as having no fast activity at onset. These findings underscore the importance of considering multiple biomarkers and further study to understand how best to apply a multifactorial approach to the analysis of presurgical EEG datasets.
Funding: NIH R01NS094399, NIH K12NS098482, AES/Epilepsy Foundation Research and Training Fellowship for Clinicians (AES 027758)