Welcome to join this lecture and learn how data science and AI technologies can help advance medical research and clinical care toward a smarter future!
📌Electronic health records (EHR) provide a rich longitudinal view of patient health, recording diagnoses, laboratory tests, medications, procedures, and other clinical encounters over time. Rather than treating these records as static covariates, this talk views them as multivariate clinical event histories. I will present a point-process perspective for two complementary learning tasks from such data. First, I will introduce a hidden graphical modeling framework for multivariate Cox processes, with the goal of learning clinical dependency networks among different types of EHR events. Second, I will discuss generalized point process additive models for patient risk prediction, where high-dimensional longitudinal event histories are used to predict scalar clinical outcomes such as mortality, readmission, disease risk, or treatment response. Together, these projects illustrate how point process models can support both biomedical knowledge discovery and patient-level prediction from EHR.
📅 Lecture Information
▪️ Topic: Learning Clinical Event Networks and Predicting Patient Risk from Electronic Health Records
▪️ Speaker: Kuang-Yao Lee (Associate Professor, Department of Statistics, Operations, and Data Science (SODS), Fox School of Business, Temple University)
▪️ Date & Time: July 28, 2026 (Tuesday), 12:10–13:30
▪️ Venue: Room IR630, 6th Floor, International Academic Research Building
▪️ Registration Link: https://forms.gle/jpR4k9goMCJh8e927
▪️ Registration Deadline: Until 12:00 PM, July 23, 2026 (Thursday); online registration only
▪️ Format: Hybrid (On-site: 60 participants; Online: unlimited)
▪️ Organizer: Biomedical Artificial Intelligence Academy of KMU, Department of Artificial Intelligence in Medicine of KMU
▪️ Note: ☕ Lunch will be provided for on-site participants. Everyone is welcome to join!
※ This lecture is eligible for KMU faculty growth credits. Participants must complete both sign-in/out and the satisfaction survey to receive credit points.
We look forward to your participation!