By Mark L. Dlugoss
Senior Contributing Editor
My first encounter with the concept of artificial intelligence (AI) was Stanley Kubrick’s science fiction film, “2001: A Space Odyssey.” When that film was released in 1968, man had not yet landed on the moon, and computers were specialized machines utilized by scientists and mathematicians. The computer’s evolution as a technology for the general public was in its early infancy.
Many movie critics saw the film as “dark and apocalyptic,” while others saw it as “optimistic hope for humanity.” One of the dark sides of the film was the character “HAL/AL 9000,”— or simply known as “HAL.”
HAL was a supercomputer programmed with the latest and greatest in AI. The computer spoke with a soft and calm voice in a simple conversational manner. It managed the functions of the spaceship and it socially engaged with the crew of the spaceship as an equal member of the team. HAL was considered a triumph in technological innovation.
Man and machine co-existed during the space mission until the crew and HAL came to a crossroads regarding the computer’s decision-making processes. One of the famous scenes from the film is the confrontation between the mission captain, Dave, and HAL. Because of HAL’s superior AI skills, the meeting did not end well for the crew.
The Looper, a website that covers pop culture, best described HAL as “this seemingly kind computer, full of authentic human emotions — who later turns out to be capable of cold, dispassionate murder — made an entire generation of filmgoers suspicious against helpful machines.”
I recall walking out of the movie theater, thinking how special it would be to interact with an intelligent piece of machinery that could help mankind deal with the complicated issues of the world. In the same thought, I wondered, could a machine really have emotions and intellect to decipher between right and wrong beyond what it was programmed to do? And if it disagreed with what it was programmed to do, could a computer really override human decisions?
At the time, my conclusions about AI were: “Hey, it’s just science fiction!” Fast forward 50-plus years, today the evolution of AI—and its cousins “deep learning” (DL) and “machine learning” (ML)—are very real. AI is making headway into the mainstream of world existence.
As we enter the new decade, AI has evolved into medicine and scientists expect it to transform the medical field in ways never thought possible. Digital-based companies (the Googles and IBMs of the world) are driving this transformation because they see the potential AI brings to medicine and healthcare. These companies have discovered the value of what DL can achieve in a clinical setting and how scientific models, when applied in medicine, can surpass and enhance the expertise of humans.
For physicians, DL has made medical imaging an easy and simple analysis and has improved the diagnosis of various medical conditions. Physicians can expect AI to play a significant role in clinical medicine and research soon.
Surprisingly, ophthalmology is one of the leaders in medicine when it comes to AI because of ophthalmic imaging. With the use of optical coherence tomography (OCT) and corneal topography, AI and DL are already here. These systems are likely to make an even greater impact in clinical ophthalmology and eye research as the decade progresses.
To get an idea of how fast AI is moving into ophthalmology, the journal Ophthalmology reported in 2020 that over 725 papers using the term “artificial intelligence” or “deep learning” appeared in the literature between 2015 and 2020, with ophthalmology publishing 10 times as many papers in 2020 compared with 2017.
As ophthalmology progresses in this arena, physicians can expect more papers to be published, paving the way for AI and DL to become more understandable and more commonplace in ophthalmic practices.
The promising areas of AI, DL, and ML in ophthalmology appear to be in retina, but the future for other ophthalmic subspecialties is also encouraging as the technology builds.
To date, researchers have developed and applied AI and DL models to ocular imaging, fundus photographs, and OCT, allowing retina specialists to detect and analyze age-related macular degeneration (AMD), diabetic retinopathy (DR), macular edema (ME), reticular pseudodrusen, and retinopathy of prematurity (ROP).
In glaucoma, ML has been utilized to analyze visual fields and to identify glaucomatous disc cupping. With these models, ophthalmologists will have the capability to screen better for glaucoma and provide support assessment as the disease progresses.
In cataract surgery, AI and ML systems can diagnose and evaluate cataracts, including pediatric cataracts. Ophthalmic surgeons can expect AI to improve and perfect intraocular lens (IOL) calculation formulas. Advanced AI models will someday provide better surgical outcomes.
In refractive surgery, researchers are developing AI and ML models to treat and evaluate corneal ectasia and keratoconus. These models also will improve the analysis of corneal topography, refractive IOL formals, surgical accuracy, refractive errors, and contrast sensitivity.
Of course, with the development of this advanced technology, many ophthalmologists have expressed concerns that AI will replace them as physicians. In actuality, AI may be more of a godsend for ophthalmologists.
Upfront, the number of ophthalmologists is declining in relation to the growing number of patients. Since many serious eye diseases occur with an aging population, the Baby Boomers will dominate the age demographic through the year 2040.
With an aging population developing cataracts, glaucoma, and many retinal diseases, there are just not enough ophthalmologists to address the growing patient load. On top of that, basic eye care will take a backseat as serious eye diseases take precedent.
AI, incorporating ML and DL, is positioned to provide the necessary assistance to narrow the manpower gap between ophthalmologists and patients. Researchers who are on the front lines in the development of AI systems believe the technology will not replace ophthalmologists but complement the clinical process.
The AI technology basically will collaborate with clinicians to enhance vision care. It will have the potential to simplify the diagnostic process, providing quicker and more efficient care for patients.
Besides, according to researchers, machines are not good at making medical decisions beyond the diagnosis. Studies on this matter have concluded that AI working with clinicians has provided better outcomes than if the machines and clinicians worked independently.
Another advantage of AI is that the technology will evolve the role of the ophthalmologist and will improve the physician-patient relationship. As outcomes and results of AI technology receive validation in clinical trials and produce the results that they claim they do, ophthalmic clinicians will come to rely on them as just another tool in their medical armamentarium.
Over the past 15 years, the physician-patient relationship has been disintegrating. Ophthalmologists are seeing more and more patients. Because of those caseloads, clinicians are spending less and less time with each patient. Many patients are leaving their appointments unsatisfied and less assured about their vision care.
AI may present an opportunity for physicians to revive the human aspect of medicine. By utilizing the results and diagnoses that AI systems offer, physicians will be able to co-manage those results for better patient outcomes. It will allow physicians to spend more time with their patients to address their questions and concerns as they apply to their vision care.
Of course, as AI becomes more entrenched in practices, patients’ expectations also will increase. Patients will expect their ophthalmologists to provide a comprehensive vision care plan. They will want more in-depth information and won’t accept simple answers to their questions. With AI, the information will be at clinicians’ fingertips. Good patient outcomes will mean better physician-patient relations in the long run.
As ophthalmologists enter a new decade, there are still plenty of questions and challenges regarding AI. However, the good news is the future of medicine will only improve as clinical-based technology helps clinicians provide better eye care in an ever-changing world.
Maybe the ophthalmic world should look at AI from the same perspective that movie critics saw the film “2001: A Space Odyssey.” Right now, AI may appear “dark and apocalyptic,” but as the technology becomes more common in ophthalmology, AI will provide an “optimistic hope for humanity.”
The Ophthalmic Project is a series of blogs and podcasts written and produced by Mark L. Dlugoss, senior contributing editor. The Ophthalmic Project will address trends, events, and observations in the world of ophthalmology. Dlugoss has nearly 25 years of experience covering the ophthalmic market.