May 31, 2018
3 min read
Save

AI may be an asset to ophthalmic surgeons

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

There are many definitions of artificial intelligence, but my favorite is: “The theory and development of computer systems able to perform tasks that normally would require human intelligence, such as visual perception, speech recognition, decision-making and translation between languages.”

The concept of AI first surfaced in 1956 in a symposium at Dartmouth College. In medicine, significant investments are being made by our largest corporations including IBM (Watson), Google and Microsoft, and many of these efforts are focused on ophthalmology. I have found that when significant capital and human intelligence are applied to any area, remarkable and many times disruptive innovation follows.

In our field, especially exciting is the analysis of images, such as the established ability of AI to grade the severity of diabetic retinopathy when analyzing a fundus photograph. In well-done studies AI performs as well as a triad of retina specialists in grading diabetic retinopathy. The value of such an application of AI through telemedicine might well bring enhanced diagnosis in the diabetes field to many patients who could not otherwise access an ophthalmologist, much less a retina specialist.

We today as a country do a poor job of monitoring our patients with diabetes for the onset of retinopathy, and we need a better model for delivery of this service. A fundus photograph taken in a primary care office and then “read” by AI might help select the patients in need of an ophthalmologist for treatment. The same would be true in monitoring other retinal pathologies including age-related macular degeneration, whether wet or dry.

Interesting to me is that AI as applied to grading diabetic retinopathy cannot make an alternative diagnosis such as branch vein occlusion, which would be readily performed by a trained ophthalmologist with human intelligence (HI). It might, for example, grade branch vein occlusion as mild background diabetic retinopathy, which could interfere with timely therapy. Another exciting application is enhancement of IOL power calculation, especially in complex eyes with prior refractive surgery or other corneal pathology.

As these AI algorithms are developed, HI will need to program AI in a way it can identify patterns that do not fit and require HI for interpretation. AI can certainly provide us a differential diagnosis for any set of symptoms or signs, and this is potentially a valuable service of AI to the clinician, reducing the need for memorization of long lists of possibilities. AI could also handicap the list, placing the most likely possibilities first. In extremely complex medical treatments, such as the use of multiple medications to treat cancer with differing responses from patients with different genetic markers, AI could be helpful in recommending the ideal treatment plan. Translation between languages will be especially helpful as the world continues to flatten and clinicians at a clinic are asked to help care for a patient in a distant country or onboard an airplane of another nationality.

PAGE BREAK

The performance of tasks is especially challenging for AI, but when coupled with robotics remains inside the possible. A computer coupled with a robot has been able to close a simple laceration successfully and in a quality fashion, but it is today hard for me to imagine a computer coupled with a robot performing cataract surgery, doing a total knee replacement or peeling an epiretinal membrane without a human involved. It is easy for me to imagine AI assisting a physician or even replacing one in the diagnosis and recommendation of appropriate therapy for many common medical conditions.

This is the double-edged sword of AI. Many fear that AI will replace jobs currently performed by humans and do them better and at a lower cost. Some predict this could lead to mass unemployment or at a minimum displacement of many workers. The most pessimistic remind us of HAL 9000, the computer in the Stanley Kubrick classic 2001: A Space Odyssey, that took over the Discovery One spacecraft and disposed of the human crew. No one to date has been able to incorporate ethics or conscience into AI. Other more optimistic observers predict that AI will lead to a golden age of greater productivity and everyone’s standard of living will be enhanced. The likeliest outcome to me is somewhere in between. I suspect AI will replace HI in some tasks while others will be spared this threat.

In our field, I expect AI to be an asset to the ophthalmic surgeon’s practice, with some rote tasks, such as performing a refraction, evaluating and grading a fundus photograph for diabetic retinopathy, or reading an OCT to evaluate the status of macular degeneration, being efficiently and effectively done by AI. Most electrocardiograms today come with a reading and diagnosis provided by AI before being seen by the physician, and the sensitivity and specificity are high. Considering the growing demand on the ophthalmic surgeon to care for our aging population combined with the growth in safe and effective surgical procedures in the face of a slowly declining cohort of ophthalmic surgeons, I see AI as an exciting positive for our field.

Disclosure: Lindstrom reports no relevant financial disclosures.