It leads to fewer medical errors than any ophthalmologist diagnosing a retinal abnormality. It leads to tremendous cost savings and faster care. It reduces the need for more ophthalmologists, potentially reversing a shortage of ophthalmologists by 2025 predicted by the Association of American Medical Colleges. What is this seemingly magical answer to health care challenges? It’s artificial intelligence (AI), and it could be coming to a community practice near you within the next decade.
In a recent editorial in JAMA, the use of AI with deep learning technology for diabetic retinopathy was discussed in light of a study in which the use of AI for diabetic retinopathy screening was described. Deep learning is a type of machine learning technology being pursued by the biggest technology companies (eg, Apple and Facebook) along with leading technologies around the United States and the rest of the world. These techniques are beginning to detect diabetic retinopathy with very high sensitivity and specificity.
The next step is a logical one: develop systems that can detect most retinal pathology without the need for a retina specialist. If such systems can be implemented, would they eliminate the need for much of the work done by a retina specialist today with respect to diagnosing and proposing management for retinal diseases?
As the editorial on this topic mentioned, would ophthalmologists trust the results of such imaging without independently viewing the images? Does the retina specialist become an information specialist, who assimilates information from deep learning technology and coordinates subsequent care?
Some experts in this area might say that such approaches in medicine are extremely likely to be incorporated during the next decade or two. Should the ophthalmology field embrace this, or look on these changes as a threat to the field?
Such advances probably should be pursued with enthusiasm. Any approach that has the possibility of reducing medical errors, improving quality of care, and providing incremental cost-effectiveness ratios that could be cost-saving sounds like a potential winner.
Yes, the systems will have to be validated, and patients will need to learn to trust a diagnosis that an information specialist concludes is correct without possibly ever having evaluated the pathology being managed. But if that’s what it takes to improve the efficiency, timeliness, and cost-effectiveness of medical care, then AI should be welcomed into the ophthalmology arena. Now, when robots also can implement the care recommended by AI . . .
Conflict of Interest Disclosure: Dr. Bressler reports a patent on a system and method for automated detection of age-related macular degeneration and other retinal abnormalities.