OHSU Informatics Faculty and Former Postdoc Develop Machine Learning System That Outperforms Experts at Diagnosing Childhood Eye Disease

An OHSU biomedical informatics faculty and a former postdoctoral fellow of the program are among a group of researchers who have developed a deep convolutional neural network system that outperformed clinical experts in the diagnosis of Retinopathy of Prematurity (ROP), a disease that causes blindness in premature infants. The results of their work have been published in the journal, JAMA Ophthalmology.

The system was tested against eight ROP experts, performing better than six of them and comparable to the other two.

The OHSU faculty and senior author of the study was Michael Chiang, MD, professor of ophthalmology and of medical informatics and clinical epidemiology at OHSU. Also an author was Jayashree Kalpathy-Cramer, PhD, associate professor of radiology at Harvard Medical School. Dr. Kalpathy-Cramer was a postdoctoral fellow in biomedical informatics at OHSU from 2006-2009.

The study was also reported on by Oregon Public Broadcasting.