When
Where
Abstract:
Modern medical imaging systems produce images through the imaging chain that consists of complicated hardware and sophisticated computational methods. Image science provides a framework for the objective assessment of image quality, which plays a critical role in exploring, evaluating, and optimizing medical imaging systems. However, because of the great number of system parameters that can affect image quality, the large variety in objects to be imaged, and ethical limitations, it is often impractical to assess imaging systems via clinical imaging trials. This can hinder the development of emerging imaging technologies. Due to these reasons, there has been growing interest in virtual imaging trials (VITs) that can emulate the clinical imaging process and permit the interpretation and analysis of medical images in silico. However, effective VITs rely on two critical components: stochastic object models (SOMs) that capture realistic anatomical variations, and figures-of-merit that relate to the ability of an observer to perform specific clinical tasks in images. Establishing realistic SOMs and computing task-based measures for such VITs present significant challenges. Machine learning methods have been actively employed to perform a wide variety of image-based tasks. In this talk, I will present an augmented generative adversarial network (GAN) method named AmbientGAN for establishing realistic SOMs from imaging measurements. I will also describe supervised learning-based methods and GAN-based Markov-Chain Monte Carlo method for approximating the Bayesian Ideal Observer for computing task-based measures of image quality. Finally, I will review our recent advancements in optimal visual search strategies and introduce a reinforcement learning method for approximating the Ideal Searcher that can be useful in understanding and improving the ways by which radiologists interpret images and make radiological decisions.
Bio:
Weimin Zhou, PhD is a Tenure-Track Assistant Professor at the Global Institute of Future Technology at Shanghai Jiao Tong University (SJTU). Before joining SJTU in 2022, he was a Postdoctoral Scholar in the Department of Psychological & Brain Sciences at the University of California, Santa Barbara (UCSB). Dr. Zhou received his Ph.D. degree in Electrical Engineering from Washington University in St. Louis (WashU) in 2020. During his Ph.D. study, he worked as a Research Assistant in the Department of Biomedical Engineering at WashU and a Visiting Scholar in the Department of Bioengineering at the University of Illinois Urbana-Champaign (UIUC). He possesses broad expertise in image science, computational image formation, visual perception, and machine learning. Dr. Zhou is the recipient of the SPIE Community Champion Award and the SPIE Medical Imaging Cum Laude Award. He serves as a program committee member for SPIE Medical Imaging and a peer reviewer for a variety of medical imaging journals.