Medical Student University of Texas Medical Branch in Galveston Galveston, Texas, United States
Disclosure(s):
Sean O'Leary: No financial relationships to disclose
Introduction: The treatment of drug-resistant epilepsy remains a significant challenge. Surgical resection or laser ablation of the epileptogenic zone (EZ) often offers the highest chance of seizure freedom. Grinenko et al. [Brain. 2018; 141(1):117–131] demonstrated that intracranial EEG (iEEG) data analysis can be used to identify a time-frequency pattern or “fingerprint” to localize the EZ. Complex algorithmic techniques, lack of standardization, and inaccessibility pose a hurdle to surgical programs that may consider fingerprinting as a tool to aid resection. R Analysis and Visualization of Intracranial EEG (RAVE) is a powerful, free, open-source, NIH-funded software designed to analyze iEEG data through a web browser on an internet-connected device. Previously we presented an accessible implementation of a fingerprinting algorithm with RAVE; here we report confirmatory testing on 10 patients.
Methods: Our fingerprint module consists of three steps, following Grinenko et al. First, iEEG seizure data is pre-processed with Morlet wavelet transform and bipolar normalization. Second, each electrode's time-frequency map is run through an image processing algorithm to extract the EZ fingerprint biomarkers: (1) pre-ictal spikes, (2) multiband fast activity, and (3) low-frequency suppression. Each electrode is ranked by biomarker and combined “fingerprint” score derived from machine learning based on work by Woolfe et al. [J Neurosci Methods. 2019; 325:108347]. Data is then displayed via time-frequency plot for each electrode, and the fingerprint score is projected across the brain with a 3D brain map viewer.
Results: Our fingerprint module in RAVE significantly streamlines EZ isolation. Preliminary data from a collective 878 electrodes across 10 patients, who were seizure-free following resection (n=5) or had laser ablation (n=5), showed EZ identification with an accuracy of 90% and specificity of 99%.
Conclusion : We incorporated an algorithm to detect the epileptogenic zone fingerprint into RAVE, allowing free distribution and use of this research algorithm to surgical epilepsy programs.