[BME Seminar] The generalist medical AI will see you now

ON2024-04-19TAG: ShanghaiTech UniversityCATEGORY: Lecture

Topic: The generalist medical AI will see you now

Speaker: Assistant Professor Pranav Rajpurkar, Department of Biomedical Informatics, Harvard University (Harvard)

Date and time: April 22, 10:00–11:30

Venue: Room 103, BME Building

Host: Cui Zhiming


Abstract:

Accurate  interpretation of medical images is crucial for disease diagnosis and  treatment, and AI has the potential to minimize errors, reduce delays,  and improve accessibility. The focal point of this presentation lies in a  grand ambition: the development of ‘Generalist Medical AI’ systems that  can closely resemble doctors in their ability to reason through a wide  range of medical tasks, incorporate multiple data modalities, and  communicate in natural language. Starting with pioneering algorithms  that have already demonstrated their potential in diagnosing diseases  from chest X-rays or electrocardiograms, matching the proficiency of  expert radiologists and cardiologists, Dr. Rajpurkar will delve into the  core challenges and advancements in the field. The discussion will  navigate towards the topic of label-efficient AI models: with a scarcity  of meticulously annotated data in healthcare, the development of AI  systems capable of learning effectively from limited labels has become a  key concern. In this vein, Dr. Rajpurkar will delve into how the  innovative use of self-supervision and pre-training methods has led to  algorithmic advancements that can perform high-level diagnostic tasks  using significantly less annotated data. Additionally, he will talk  about initiatives in data curation, human-AI collaboration, and the  creation of open benchmarks to evaluate the generalizability of medical  AI algorithms. In sum, this talk aims to deliver a comprehensive picture  of the state of ‘Generalist Medical AI’, with the advancements made,  the challenges faced, and the prospects lying ahead.


Biography:

Pranav  Rajpurkar is an Assistant Professor at the Department of Biomedical  Informatics, Harvard University. His research focuses on developing AI  systems that can interpret medical data, reason through complex  problems, and communicate at an expert level, with the goal of creating  AI doctors that can work independently or alongside human physicians.  Rajpurkar has published over 100 academic articles in journals like Nature, NEJM, and Nature Medicine,  garnering more than 25,000 citations in prestigious journals. He has  been recognized with numerous awards, including Forbes 30 Under 30 in  science in 2022, MIT Tech Review’s Innovator Under 35 in 2023, and  Google Research Scholar.