Search Papers | Poster Sessions | All Posters
Poster A60 in Poster Session A - Tuesday, August 6, 2024, 4:15 – 6:15 pm, Johnson Ice Rink
Fast & Accurate Classification of Task Stages in ECoG Generalizes to Continues Recordings
Amirhossein Khalilian-Gourtani1, Yasamin Esmaeili2, Adeen Flinker1; 1Neurology Department, New York University School on Medicine, 2Biomedical Engineering Department, New York University Tandon School of Engineering
In this study, we explore the decoding of task-specific cognitive states and how they generalize to continuous neural recordings. Using electrocorticography (ECoG) data from eight neurosurgical patients performing various speech production tasks, we compare the performance of three time-series classification models (predicting perception, speech and rest). Results demonstrate that our proposed framework based on a mini-rocket model achieves highest accuracy and fastest inference time compared to the other models. Further, we assess the electrode importance for decoding and find a strong correlation between signal activity and decoding accuracy across sensory and motor regions. However, we do not find this relationship in other frontal regions. Our framework exhibits robust generalization capability across recording sessions, showcasing its potential for analyzing continuous neural recordings and deciphering cognitive states accurately.
Keywords: electrocorticography speech perception and production classification decoding