Dr. Tony Liang

Abstract: Neural networks have revolutionized AI, but a critical question persists, especially in healthcare: can we truly trust them? While performance gains have been rapid, issues like miscalibration, poor generalization, inherent bias, and unstable feature learning remain significant hurdles. Imagine a miscalibrated cancer diagnosis model confidently misclassifying a malignant tumor-–the consequences are devastating. This talk will discuss some of the root causes of these trustworthiness and safety concerns within the healthcare context and introduce a novel approach to enhance the reliability of neural network-powered AI.

Biography:

Dr. Gongbo “Tony” Liang, Assistant Professor of Computer Science at Texas A&M University-San Antonio, is dedicated to building trustworthy AI for medical imaging. His research addresses the crucial challenges of reliability in neural networks used for disease diagnosis, covering a wide range of imaging modalities and conditions like breast cancer, lung cancer, Alzheimer's disease, stroke, and bone fractures. Since 2017, he has published over 25 papers on this topic and developed a breast cancer diagnosis software now licensed for industrial use.