The medical data housed in patient records are a clinical researcher’s dream: the key, potentially, to better tools to treat disease and screen with precision. They’re also a computer scientist’s nightmare: locked away in hospital systems, subject to restrictive data-sharing agreements, and often too messy to make use of.
A new open science project wants to accelerate ethical AI in medicine by doing the hard work of collecting and cleaning that data. Nightingale Open Science launched in December with $6 million in funding, led by Schmidt Futures, the philanthropy of ex-Google CEO Eric Schmidt and his wife, Wendy Schmidt. (It has no affiliation with Google’s controversial health record-mining partnership with Ascension, which went by the code name Project Nightingale). It will freely share de-identified clinical datasets with researchers, linking medical images like X-rays, ECG results, and biopsy slides — 40 terabytes worth, to start — to outcomes from partnered health systems. Hundreds of researchers have signed up for access in its first month.
The nonprofit, co-founded by UC Berkeley ER physician and machine learning researcher Ziad Obermeyer and University of Chicago computation and behavioral science researcher Sendhil Mullainathan, also aims to shift the way researchers think about applying artificial intelligence to medicine. Instead of aiming to reproduce clinical judgement with algorithms, each of the initial datasets are structured like a problem set, including data about a medical mystery that has stymied doctors — but may well be answerable by machine learning algorithms that see medical imaging differently. “It’s really like there’s this alien intelligence looking at this picture in totally different ways from the way a human looks at it,” Obermeyer said.
This article is exclusive to STAT+ subscribers
Unlock this article — and get additional analysis of the technologies disrupting health care — by subscribing to STAT+.
Already have an account? Log in
About the Author Reprints
To submit a correction request, please visit our Contact Us page.
