Chatting About Pathology

Statistical Bioinformatics Seminar

Speaker: Dr Drew Williamson (Emory University)

Abstract: Generative models such as ChatGPT have catapulted AI further into the mainstream than ever before, while interested parties in healthcare have begun to test these models in medical settings. However, these models are rarely purpose-built for medical applications and the majority of those that do exist are targeted at a broader segment of healthcare than any individual provider practices. I’ll discuss the design, training, and evaluation of PathChat, our multimodal generative model created specifically for pathology. I’ll also discuss possible future directions in this line of inquiry.

About the speaker: Drew Williamson completed a BA in Mathematics at Oberlin College, an MD at Case Western Reserve University, residency in Anatomic Pathology at Brigham & Women’s Hospital, and fellowships in Molecular Genetic Pathology and Clinical Informatics at Mass General Brigham. This training was followed by postdoctoral research in the lab of Faisal Mahmood, PhD at Brigham & Women’s Hospital and Harvard Medical School. A board certified pathologist, Drew’s research focuses on the applications of deep learning-based methods to pathology data, from histology images to genomics to, most recently, natural language. He is now an Assistant Professor in the department of Pathology & Laboratory Medicine at Emory University.