AIMI Research Meeting: Diagnose Like a Doctor: Integrating Domain Knowledge into Medical AI - Yuyin Zhou, PhD

Abstract:
Deep learning has largely advanced the field of healthcare and medicine by offering an avenue to deliver automated medical image analysis. Nonetheless, existing deep learning systems generally focus on a less structured approach in which medical representations are not explicitly modeled as in real clinical diagnoses, posing a challenge for interpretability and trustworthiness of such models. 

Based on these challenges, in this talk, I will first discuss how to make medical AI systems approach real clinical expertise by embedding different types of domain knowledge. Next, I will showcase the benefits of exploiting medical knowledge for learning with non-standardized datasets, mitigating dataset biases, and facilitating knowledge transfer across clinical environments, thus leading to more verifiable, reliable, and practical AI solutions. Finally, I will touch on the pervasiveness of medical knowledge in real-world clinical applications and identify other important future directions for furthering medical AI.

About:
Dr. Yuyin Zhou is a postdoctoral researcher at Stanford University. She received her Ph.D. from the Computer Science Department at the Johns Hopkins University in 2020. Yuyin’s research interests span the fields of medical image computing, computer vision, and machine learning, especially the intersection of them. Her project with Johns Hopkins Medicine on organ segmentation has been featured on National Public Radio. She has over 20 peer-reviewed publications at top-tier conferences and journals including CVPR, ICCV, AAAI, Medical Image Analysis, etc. She served as a senior program committee at IJCAI.