SEEG-based Localization of the Epileptogenic Zone from Complexity Measures Using Machine Learning

Poster Session F - Tuesday, April 16, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Yorguin Mantilla^1,2,3 (yjmantilla@gmail.com), Jian Li⁴, Dileep Nair⁵, Karim Jerbi³, Richard Leahy⁶; ¹Grupo Neuropsicología y Conducta (GRUNECO), Facultad de Medicina, Universidad de Antioquia, Medellín, Colombia, ²Semillero de Investigación NeuroCo, Universidad de Antioquia, Facultad de Medicina & Facultad de Ingeniería, Universidad de Antioquia, Medellín, Colombia, ³Cognitive and Computational Neuroscience Laboratory (CoCo Lab), University of Montreal, Quebec, Canada, ⁴Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Charlestown, MA, USA; Center for Neurotechnology and Neurorecovery, Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA, ⁵Charles Shor Epilepsy Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA, ⁶Signal and Image Processing Institute, University of Southern California, Los Angeles, United States

Introduction: In this research, we sought to delineate the epileptogenic zone using a dataset from the Cleveland Clinic, encompassing 28 patients who successfully underwent resective surgery and had prior SEEG recordings from both ictal and interictal periods. Methods: From time-windowed segments of these recordings, we derived complexity features and characterized them using their mean and standard deviation. Our analysis incorporated features such as Lempel-Ziv complexity, various entropies, fractal dimensions, and the 1/f slope of the brain activity spectrum, among others. We trained three distinct Logistic Regression Models: one using only ictal data, another using only interictal data, and a hybrid model leveraging both periods. Results: Our findings underscored that while the interictal period might be less informative in isolation, it enhances the insights drawn from the ictal phase when combined. A pivotal aspect of our research was discerning a distinctive epileptogenic zone fingerprint. Feature importance analysis pinpointed the Mean Lempel-Ziv Complexity, the standard deviation of the 1/f Slope, and the standard deviation of specific fractal dimensions as the most significant characteristics differentiating resected locations. Conclusion: These results not only contribute to understanding the epileptogenic zone but also foster discussions about complexity in the brain, particularly in the context of the brain criticality hypothesis.

Topic Area: OTHER