Machine Learning and Epilepsy: Classifying Epileptic Brain States Using Sugihara Causality Networks

Abstract

Epilepsy is a brain disease that profoundly affects the personal lives of its patients as well as the health care system needed to address it. Machine learning is a field of computer science experiencing a renaissance period at a time where computers can do more work in less time than ever before. While industry has moved rapidly towards implementation of machine learning in profitable and futuristic endeavors like autonomous driving and image recognition and creation, some basic and essential societal problems remain relatively unexamined through the lens of machine learning research, an example of which is epilepsy. This work attempts to start to fill this gap and shed light on the ways machine learning can address complex and intricate problems like epilepsy. Through it, several machine learning algorithms will be surveyed, and some simple implementations and tutorials be provided through an external companion. Furthermore, a novel definition of brain communication is introduced in the form of causality, which will also be discussed more broadly. This definition is then used to construct a network of brain communication patterns in the context of which the author attempts to apply machine learning. Finally, discoveries made in this work will highlight a lack of a general understanding and framework of brain communication patterns and the magnitude and scope of future work available in the intersection of neuroscience and machine learning.