The studies, both led by researchers out of MIT’s Computer Science and Artificial Intelligence Laboratory, analyzed EHR data from patients in intensive care units to inform physician decisions and predictions related to patient outcomes.
For one study approach, called “ICU Intervene,” PhD students Harini Suresh and Marzyeh Ghassemi; undergraduate student Nathan Hunt; postdoctoral student Alistair Johnson; and researchers Leo Anthony Celi, MD, and Peter Szolovits, PhD, used deep learning to learn from past cases and make treatment suggestions for critical care. The system also explained the reasons behind its recommendations.
Another team developed a machine-learning approach called “EHR Model Transfer.” PhD students Jen Gong and Tristan Naumann led the project with Dr. Szolovits and researcher John Guttag, PhD. The model exploits ICU data to predict patient mortality and prolonged length of stay. The approach also uses natural language processing technology to align clinical concepts that are coded differently across EHR platforms, enabling data analysis and prediction at various health systems.
More articles on artificial intelligence:
Can AI improve cancer care in remote areas? 3 questions with Dr. Andrew Norden of IBM Watson Health
6 thoughts on population health with Allscripts Analytics CMO Dr. Fatima Paruk
Paper records are here to stay: 4 questions with Illinois Pain Institute’s Dr. John Prunskis
