EMBARGOED FOR RELEASE: 10:30 P.M. (ET), WEDNESDAY, MAY 2, 2018
Want to embed a link to this study in your story? Link will be live at the embargo time http://jamanetwork.com/journals/jamaophthalmology/fullarticle/10.1001/jamaophthalmol.2018.1934
Bottom Line: An algorithm could detect signs of a serious eye disease in images from premature infants with accuracy comparable to or better than human experts.
Why The Research Is Interesting: Retinopathy of prematurity (ROP) is a leading cause of childhood blindness worldwide. The decision to treat is primarily based on the presence of plus disease, which is when retinal vessels are dilated and twisted. However, clinical diagnosis of plus disease can be highly subjective and variable.
What and When: A machine learning algorithm was trained to diagnose plus disease using 5,511 retinal photographs. Data were collected from July 2011 to December 2016 and analyzed from December 2016 to September 2017.
Study Measures: The algorithm to detect plus disease was tested on an independent set of 100 images against eight ROP experts.
Authors: Michael F. Chiang, M.D., Oregon Health and Science University, Portland, Jayashree Kalpathy-Cramer, Ph.D., Massachusetts General Hospital, Boston, and coauthors
Results: The algorithm diagnosed plus disease with comparable or better accuracy than human ROP experts.
Study Limitations: Algorithms in artificial neural networks are only as good as the data on which they are trained. It is unknown how factors such as image quality, resolution, different camera systems and field of view may affect the output of these deep learning systems.
For more details and to read the full study, please visit the For The Media website.
Editor’s Note: Please see the article for additional information, including other authors, author contributions and affiliations, financial disclosures, funding and support, etc.
# # #
For more information, contact JAMA Network Media Relations at 312-464-JAMA (5262) or email firstname.lastname@example.org.