Abstract

Digital health is not new, but the technologies that enable it are evolving rapidly. Uncritical adoption of emerging tools can waste resources and, worse, cause harm. We argue for choosing the right tool for the right job, illustrated through real-world AI applications in healthcare. Specifically, we show that traditional data mining methods outperform large language models (LLMs) for early-stage sarcopenia detection. We then demonstrate how LLM-based clinical decision support systems become more reliable when augmented with conventional search and retrieval. Finally, we overview a recently deployed, non-invasive cancer prescreening system built with LLM technology. No medical background is assumed or required.

About the Speaker

Ophir Frieder focuses on scalable information processing systems with particular emphasis on health informatics. An inductee of the Florida Inventors Hall of Fame and a Member of both Academia Europaea and the European Academy of Sciences and Arts, he is a Fellow of the AAAS, ACM, AIMBE, IEEE, and NAI, and an Inaugural Member of both the ACM SIGIR Academy and the WIC Web Intelligence Academy. He co-founded multiple companies, advised or consulted for many others, and is a named inventor on over 150 issued US patents and 50 pending applications, a significant portion of which are in products or systems in daily use. He is a member of the Computer Science faculty at Georgetown University and the Biostatistics, Bioinformatics, and Biomathematics faculty at the Georgetown University Medical Center.