Artificial intelligence holds the promise of diagnosing eye diseases faster and more accuracy than physicians. It is possible that technology could replace some of the more routine eye examinations hat physicians perform. While this may be the case, a new study indicates that the most effective application of advanced technology is with physicians and algorithms working in unison to track and detect eye diseases.
The research builds upon developments from Google AI, which had shown that Google’s health algorithm works almost as well as human medics when screening patients for the common diabetic eye disease called diabetic retinopathy (retinal vascular disease). The new research sought to inquire whether the algorithm could do more than simply diagnose disease.
This led the researchers to construct a computer-assisted system capable of “explaining” the algorithm’s diagnosis. The outcome was that the modified system could improve upon an ophthalmologist’s diagnostic accuracy; furthermore, by working in tandem with an ophthalmologist, this ‘human-machine’ partnership was also shown to improve upon the algorithm’s accuracy.
In trails, ten ophthalmologists were requested to read different eye disease images under one of three conditions: unassisted, grades only (assisted by the algorithm), and grades + heatmap (assisted by the algorithm with additional analysis). The outcome was that both types of assistance improved physicians’ diagnostic accuracy.
Lead researcher, Dr. Rory Sayres said: “What we found is that AI can do more than simply automate eye screening, it can assist physicians in more accurately diagnosing diabetic retinopathy. AI and physicians working together can be more accurate than either alone.”
The research has been published in the journal Ophthalmology. The peer-reviewed paper is titled “Using a Deep Learning Algorithm and Integrated Gradients Explanation to Assist Grading for Diabetic Retinopathy.”
