Speaker

Dr. Soumyanil Banerjee, Postdoctoral Research Fellow, University of Michigan

Abstract

Medical imaging and healthcare data analysis in clinical settings face significant challenges, including handling complex multi-modal data, addressing missing information, and ensuring accurate and efficient models to support critical decision-making processes. Medical images are often high-dimensional, heterogeneous, and contain subtle yet critical features that are difficult for human experts to consistently identify. Electronic health records contain a wealth of information about patients' medical histories, treatments, and outcomes, but making sense of this data to derive actionable knowledge remains challenging. In recent years, deep learning has shown immense potential in overcoming these challenges by providing robust solutions for tasks such as classification, forecasting, generation and segmentation.

In this talk, I will present several novel deep learning architectures we have developed to address key challenges in medical imaging and healthcare informatics. I will discuss how we model non-local dependencies in brain connectome data, enabling more precise localization of epileptogenic regions for pediatric patients. I would also describe how we handle missing MRI sequences through cross-sequence distillation for epileptogenic region localization. Then, I will describe a spatial-temporal graph transformer approach that we developed to forecast the evolution of COVID-19 spread. I will also share our work on enhancing medical image segmentation with dual self-distillation and conditional diffusion models, for effective radiation treatment planning. These innovations demonstrate the potential of deep learning to address critical clinical challenges and pave the way for more effective and accessible healthcare solutions. 

Bio

Soumyanil Banerjee is currently a Postdoctoral Research Fellow in the Department of Neurosurgery at Michigan Medicine, University of Michigan Ann Arbor. He received his Ph.D. degree in Computer Science from Wayne State University in 2024, the M.S. degree in Electrical and Computer Engineering from University of Michigan Ann Arbor in 2015 and the B.S. degree in Electronics and Communications Engineering from West Bengal University of Technology in 2012. He works at the intersection of deep learning, medical imaging and healthcare data analysis. His research focuses on designing novel deep learning algorithms to identify intricate patterns in real-world medical imaging and healthcare data which could facilitate treatment planning and improve patient outcomes. He has published more than 20 papers in top conferences and journals including IEEE TMI, Smart Health, Human Brain.