Knowledge Guided AI for Healthcare Analytics

AI has been rigorously explored in healthcare in recent years. Most of the AI models for healthcare are learned from the EHR data where data missingness and scarcity is often unavoidable. To achieve clinically accurate AI, it is desirable to properly incorporate prior knowledge for the model learning. In this talk, I will present some recent works of my research group where explicit and implicit knowledge from medical ontologies and LLMs are leveraged to guide the learning for analytic applications like diagnosis prediction and radiology report generation.