Deep Learning Applications in Ophthalmology

Ehsan Rahimy


Curr Opin Ophthalmol. 2018;29(3):254-260. 

In This Article

The Road Ahead

Physician-assisted automated interpretation of images in ophthalmology may eventually help improve workflow efficiency at the clinic level, allowing for more direct patient interaction. Outside the clinic, deep learning platforms appear poised to make inroads into telemedicine, on a global as well as domestic scale. For example, 30–50% of patients with diabetes do not adhere to guidelines recommending routine eye examinations to detect for retinopathy.[39,40] Potential benefits of deep learning-based screening programs would include: increased efficiency and coverage (i.e. algorithms are programmed to withstand repetitive image processing, can work in parallel, and do not fatigue), reducing barriers to access for areas where an eye care provider may not be present, providing earlier detection of referable eye disease, and decreasing overall healthcare costs through earlier intervention of treatable disease rather than resorting to more costly interventions in the more advanced phases of disease.

Looking further into the future, deep learning offers the potential to help solve a number of our overburdened healthcare system's growing problems. As of now, these algorithms have been mostly used for the detection and diagnosis of disease. However, as efforts grow towards developing datasets over an extended period of time from the same patients, could deep learning start to infer patterns of disease progression, and potentially make predictions off of them? If those images could then be tied in with systemic data points (i.e. blood pressure, hemoglobin A1c, renal function, etc.) from the corresponding patients, could it infer more comprehensive information, such as the risk of systemic morbidity/mortality? In this emerging world of precision medicine, we may one day be able to tailor treatments and intervention to those at the highest risk of disease progression at an earlier state. For example, diabetic retinopathy, could potentially be reclassified along a scale where a numeric grade denotes a patient's risk of developing DME or progressing to proliferative disease.