Since Beth Israel Deaconess Medical Center in Boston launched the MIMIC compendium of de-identified medical records in 2009, the database has proved essential to the advancement of artificial intelligence research in medicine, STAT reports.
MIMIC is free to use, with more than 12,000 researchers from around the world granted access to it in the last decade. The database is estimated to have been used in more than 500 research papers and presentations, and is widely considered to be the only open-source dataset comprehensive enough to aid in such advanced research, despite unavoidable flaws such as occasionally incomplete data and the inherent bias due to its being sourced from just one hospital.
MIMIC's patient data includes minute-by-minute vital signs, lab test results, physician notes, billing information, demographics, medication history and more, gathered from more than 46,000 patients in Beth Israel Deaconess' intensive care units. Its next update, MIMIC-IV, scheduled for a 2020 release, will contain data from around 60,000 admissions.
"I don't think that the machine learning for health community would have expanded and progressed as quickly as it has without MIMIC as a resource," Marzyeh Ghassemi, PhD, an assistant professor at the University of Toronto, told STAT.
More articles about AI:
VA appoints 1st director of AI, plans to expand AI research: 4 notes
6 hospital applications for machine learning: algorithms to predict patient violence, HIV risk & more
Viewpoint: The AI revolution will leave us 'struggling to understand'