T105. A novel method for ECOG-based localization of function

Abstract

Introduction Identifying indispensable eloquent cortex is an integral process of the surgical evaluations. The current accepted ‘gold’ standard of functional mapping of eloquent cortex is electrical cortical stimulation (ECS) suffers from known limitations. In recent years, electrocorticographic (ECOG) recording of task related high gamma (70–120 Hz) activity has emerged as a potential futuristic tool for functional brain mapping. We have shown that in case of extensive spatial sampling, there is a higher chance of accidental detection of high frequency oscillations/gamma activity over the occipital-junctional and the rolandic regions using commonly-employed statistical methods. These methods are often based on Fourier Transform which is computationally-effective but is susceptible to spectral leakage and may not differentiate between increase in gamma due to hyper-synchronization (i.e. epileptiform discharges) vs. de-synchronization (i.e. functional activation). The qualitative binary - coefficient of determination -based methods and auto-regressive models, and while are computationally-effective, discount the absolute energy altogether and do not always differentiate between contacts that exhibit statistically significant increase vs. decrease in gamma in relation to task as they are calculated as a squared value by design. In this pilot study we propose and implement a novel method for ECOG-based mapping of function and we demonstrate the enhancing effect on predictive values. Methods We studied the sensitivity, specificity, negative and positive predictive values of the novel method and compared that to commonly employed algorithms and the results of standard electrical cortical stimulation. Results We report the preliminary results from 4 motor and sensory tasks. The novel method achieved Sensitivity 99%, Specificity 97%, positive and negative predictive values of 44%, and 99% respectively, despite the extensive spatial sampling and compared favorably to commonly employed task-related gamma detection methods, specifically it minimized ’false detections’ from the occipital, perirolandic and epileptic regions. Conclusion This work focused on optimizing the current detection methods and modulate the predictive values for practical clinical use. Future prospective studies are warranted to investigate the performance of the method/detector.