Google AI detects anemia from retinal imaging data

A deep learning model developed by Google Health researchers assists providers in diagnosing anemia based on noninvasive retinal screening, rather than a traditional blood test, a recent study found.

In the study, published in Nature Biomedical Engineering, the Google Health team developed the artificial intelligence algorithm and applied it to more than 114,000 retinal fundus images from about 57,000 participants. When combined with patient data such as age and sex, the model was able to detect anemia with 88 percent accuracy.

Though the deep learning model was initially developed using a dataset of primarily Caucasian participants, to battle the ongoing issue of algorithmic bias, it was validated on a separate dataset from Asia and achieved comparable results, according to a blog post by Akinori Mitani, MD, PhD, lead author of the study and a research engineer at Google Health.

Additionally, upon discovering these quantifiable effects of anemia on the eye, the researchers conducted an analysis to find that the optic disc and surrounding blood vessels are most likely to contain signs of anemia.

"This method to noninvasively screen for anemia could add value to existing diabetic eye disease screening programs, or support an anemia screening that would be quicker and easier than a blood test," Dr. Mitani concluded. "We hope this will inspire additional research to reveal new scientific insights from existing medical tests, and to help improve early interventions and health outcomes."

More articles on AI:
Microsoft launches $40M AI for Health initiative
One-third of hospitals now use imaging AI, survey finds: 3 things to know
FDA clears Eko's AI-powered stethoscope for AFib, heart murmur detection

© Copyright ASC COMMUNICATIONS 2020. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.

 

Featured Webinars

Featured Whitepapers