Abstracts

Rationale: Electrocorticography-based functional language mapping (ECoG-FLM) utilizes ECoG signal paired with simultaneous language task presentation to create functional maps of eloquent language cortex in patients selected for resective epilepsy or tumor surgery [1-4]. However, at present, the concordance of these functional maps derived by ECoG-FLM and electrical cortical stimulation mapping (ESM) remains rather low (sensitivity 62% and specificity 75%) [for review, see 5]. These limitations in concordance impede the transition of ECoG-FLM into independent functional mapping modality. As ESM is considered a gold standard of functional mapping, we aimed to use it, for the first time, in combination with machine learning (ML) approaches ("ESM-ML guide"), to improve accuracy of ECoG-FLM. Methods: The ECoG data were collected from 11 patients (26.4±11.3 yrs; 15-52 yrs; 7 males, 4 females). Patients’ ECoG activity was recorded (g.USBamp, g.tec, Austria) during administration of language tasks [6]. Data analysis (Fig. 1): (1) All ECoG sites were divided into ESM positive (ESM (+)) and ESM negative (ESM(-)); (2) Features of ESM(+) and ESM(-) sites in ECoG signal were determined by analyzing the signal in the frequency domain; (3) ML classifiers (Random Forest (RF) and Deep Learning (DL)-based) were trained to identify these features in language-related ECoG activity; (4) Accuracy of ESM-guided classification results was compared with the accuracy of conventional ECoG-FLM analysis. Results: Preliminary results are presented. Conventional approach demonstrated: Accuracy: 58%; sensitivity: 22%; specificity: 78%.  "ESM-ML guide" approach with RF classifier demonstrated: Accuracy: 76.2%; sensitivity: 73.6; specificity: 78.78. DL-based classifier achieved the highest performances compared to all others with 83% accuracy, 84% sensitivity and 83% specificity. Comparison between single frequency bands and the whole frequency spectrum revealed higher sensitivity and specificity values when the latter was utilized (Fig. 2). Conclusions: ECoG-FLM accuracy can be improved by using "ESM-ML guide", making feasible the use of ECoG-FLM as a stand-alone methodology. Continuous (0-350 Hz) frequency analysis approach might be more favorable than a single band one, when extracting ESM(+) and ESM(-) features. Increasing sample size and finding more advantageous ML classification approaches are recommended. The long-term goal is to create a tool-box with "ready to use ESM-ML guide" algorithm trained to provide high accuracy ECoG-FLM results by classifying between ESM(+) and ESM(-) contacts in prospective sets of language-related ECoG data. Its implementation will contribute towards improved surgical outcomes.References1. Epilepsia, 2017. 58(4): p. 663-673.2. J Clin Neurophysiol, 2015. 32(3): p. e12-22.3. Clin EEG Neurosci, 2014. 45(3): p. 205-11.4. J Neurosurg, 2016. 125(6): p. 1580-1588.5. J Pediatr Epilepsy, 2015. 04(04): p. 184-206.6. J Neurosurg Pediatr, 2014. 14(3): p. 287-95. Funding: The Central Florida Health Research (CFHR) grant (PIs: Drs. M. Korostenskaja and U. Bagci).