EMBARGOED FOR RELEASE: 11 A.M. (ET), TUESDAY, DECEMBER 12, 2017
Want to embed a link to this report in your story? Link will be live at the embargo time http://jamanetwork.com/journals/jama/fullarticle/10.1001/jama.2017.18152
Bottom Line: A computing system with artificial intelligence that can learn to do tasks that normally require human intelligence could detect retinal images that did and did not show diabetic retinopathy and related eye diseases in multiethnic populations.
Why The Research Is Interesting: Diabetic retinopathy is a vision-threatening eye disease. One of the challenges of screening for diabetic retinopathy is the lack of trained individuals to assess retinal images.
Who: 494,661 retinal images from Chinese, Indian, Malay, Hispanic, African-American and White patients.
What (Study Measures): To test the performance of a deep learning computing system that was developed and trained to classify retinal images to detect diabetic retinopathy, possible glaucoma and age-related macular degeneration and compare it with human evaluators of the images.
Authors: Tien Yin Wong, M.D., Ph.D., of the Singapore National Eye Center, Singapore, and coauthors
Results: The computing system had high rates of correctly identifying retinal images with and without diabetic retinopahy and related eye diseases.
Study Limitations: Improvements could be made in the data sets used to train and test the computing system.
Study Conclusions: More research is necessary to evaluate how such a computing system could be used in health care settings to improve vision outcomes.
The following related elements also are available on the For The Media website:
- The study, “Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer,” by Babak Ehteshami Bejnordi, M.S., and colleagues.
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 email@example.com.