The electroencephalogram (EEG) is an important neuroimaging modality to measure neuronal activity of the human brain. It has its undisputed value in clinical diagnosis, particularly in the identification of epilepsy and sleep disorders and in the evaluation of dysfunctions in sensory transmission pathways. EEG Source Imaging refers to the process of reconstructing the brain source signal based on the scalp EEG data. I will present a new insight to solve brain imaging problems based on the story of graph structure of the data, for the data, and from the data. I will first discuss how to solve the EEG Source Imaging (ESI) problem using a hierarchical graphical model with a spanning tree structure constraint connected from latent landmarks source activation patterns. Next, I will talk about how to use Laplacian graph regularization to solve the supervised ESI problem. Then, I will talk about how to find common connectivity pattern and discriminative connectivity patterns from the estimated brain networks during the sleep onset period (SOP) and rapid eye movement (REM) sleep, using a multi-tasking sparse learning framework.
Dr. Feng Liu is an Assistant Professor at the School of Systems and Enterprises at Stevens Institute of Technology. Dr. Liu was a Postdoctoral Research Fellow at Harvard Medical School from 2018 to 2020. He was a research affiliate at Picower Institute for Learning and Memory at MIT and Martinos Center for Biomedical Imaging at Massachusetts General Hospital from 2018 to 2020. Dr. Liu received his Ph.D. degree from the University of Texas at Arlington in Industrial Engineering in 2018. His research interests include brain imaging, inverse problem, health informatics, machine learning, and dynamic system. Prof. Liu is the winner of the Best Paper Award at 11th International Conference of Brain Informatics in 2018, and the winner of the Best Paper Award of INFORMS Data Analytics Society in 2019.