Improving Medical Decision Making Through Observational Supervision

Collaborators: Morteza Noshad, Jared Dunnmon, Khaled Saab, Jonathan Chen, Wui Ip, Daniel Rubin

Advised by: Daniel Rubin, MD, MS

Professor of Biomedical Data Science, Radiology, and Medicine (Biomedical Informatics)

and (by courtesy) Computer Science and Ophthalmology,

Stanford University

Advised by: Wui Ip, MD

Clinical Assistant Professor, Pediatrics

Stanford School of Medicine

There is a dearth of large, well-labelled datasets in the medical community. Even when such labelled datasets are painstakingly curated, the labels are highly simplistic and task oriented, failing to generalize. What if, instead, just observing physicians interact with data in routine practice allows continuous collection of weak training labels that can be used to train models? Our project on observational supervision is exactly that!

We developed methods to extract meaningful information from observational metadata such as clickstreams and access logs in combination with clinical EHR to aid medical decision making. Check out our publications for more details.