The eyes may be well known as the, “window to the soul,” but an appreciation for the information contained within them is just beginning to dawn in our space. The images that retina specialists pore through daily are untapped treasure troves, holding far more potential than previously predicted. 

In 2016, Google published a deep-learning algorithm capable of detecting diabetic retinopathy as accurately as trained ophthalmologists. Such data seems predictable now, and authors have described similar systems for interpreting skin lesions in dermatology. 

Since then, the field has moved quickly. As Elon Musk recently surmised, “The pace of progress in AI is incredibly fast … you have no idea how fast.”1 In 2018, Google again advanced the AI ball in retina and demonstrated that deep learning can extract information “not previously thought to be present or quantifiable in retinal images,” from a single fundus image.2 It used isolated fundus images to deduce a subject’s age, gender, blood pressure, smoking status and history of a major adverse cardiac event with remarkable accuracy. It appears a deep-learning system can tell if a fundus image belongs to a male or a female with astonishing consistency. Extrapolating from such deductive abilities hints at an incredible opportunity for retinal imaging to serve as a biomarker for systemic diseases. 

On the whole, retina specialists seem vaguely aware that advanced machine-learning platforms have impacted medical fields such as radiology and cardiology, but many in our field think the technology is not yet ready to penetrate it.

On page 18, Ehsan Rahimy, MD, astutely outlines the current landscape of AI within retina and predicts where we’re headed. In the near future, such advances promise to improve access to disease screening and augment our clinical capabilities through improved prognostication and more specific application of precision medicine. 

AI will also extend into our clinics beyond image interpretation. Consider AI systems serving as medical scribes, both for patient intake and for improved physician documentation. Imagine a system that continuously improves based on your direct feedback until it is incredibly fine-tuned to your voice and documentation desires. It could alleviate administrative costs and burden tremendously, and greatly enhance the quality of the exam and discussion documentation. 

As our profession continues to evolve, I see a bright future for AI in retina. I encourage you to engage with me in creating this future for the betterment of our patients.  

REFERENCES

1. Marr B. 28 best quotes about artificial intelligence. Forbes. July 25, 2017. Available at: https://www.forbes.com/sites/bernardmarr/2017/07/25/28-best-quotes-about-artificial-intelligence/#5247b3664a6f. Accessed October 11, 2018.

2. Poplin R, Varadarajan AV, Blumer K, et al. Prediction of cardiovascular risk factors from fundus photographs via deep learning. Nat Biomed Engineer. 2108;2:158-164.