Machine Learning Interpretability Methods to Characterize Brain Network Dynamics in Epilepsy.

The rapid adoption of machine learning (ML) algorithms in medical disciplines has raised concerns about trust and the lack of understanding of their results. Efforts are being made to develop more interpretable models and establish guidelines for transparency and ethical use, ensuring the responsible integration of machine learning in healthcare. In this study we use two ML interpretability methods to gain insights into the dynamics of brain network interactions in epilepsy, a neurological disorder increasingly viewed as a network disorder that affects more than 60 million persons worldwide. Using high-resolution intracranial electroencephalogram (EEG) recordings from a cohort of 16 patients, combined with high-accuracy ML algorithms, we classify EEG recordings into binary classes of seizure and non-seizure, as well as multiple classes corresponding to different phases of a seizure. This study demonstrates, for the first time, that ML interpretability methods can provide new insights into the dynamics of aberrant brain networks in neurological disorders such as epilepsy. Moreover, we show that interpretability methods can effectively identify key brain regions and network connections involved in the disruptions of brain networks, such as those that occur during seizure events. These findings emphasize the importance of continued research into the integration of ML algorithms and interpretability methods in medical disciplines and enable the discovery of new insights into the dynamics of aberrant brain networks in epilepsy patient.