Evolution of AI in ophthalmology

Throughout history, all medical decision-making has relied upon anecdote, expert opinion, and cumbersome and limited clinical trials with delayed or nonexistent post hoc analysis and validation.

More recently with the advent of large, connected datasets and AI, the exciting potential has arisen to develop methods to optimize and validate medical decision-making. These AI approaches can be applied universally across medicine, and there is a particularly high expertise currently focused on ophthalmology.

AI and machine learning have had an impact on all sectors of our society, and medicine is certainly at the forefront because it can so greatly benefit from advances in this field. The progress that has been made in the last 10 years since we realized its power has been astronomical. AI has become important in image analysis, predictive algorithms and guidance across all fields of medicine.

AI in ophthalmology at this point is used for screening a diagnosis of diseases such as glaucoma, diabetic retinopathy, macular degeneration, cataract and keratoconus, among others. Our journey at Advanced Euclidean Solutions started about 10 years ago when we realized the ultimate power that would come and surpass any human ability to write an algorithm or formula for almost any medical mathematical problem. Our vision went beyond just predicting an outcome but guiding the physician to better future outcomes. In addition, we believed then and still do now that adjustments of practice patterns are best done with deep learning to make incremental changes to a field of medicine.

Our vision back then was that AI can do more than only predict something, but rather it can take it one step further to guide the next step and continually improve in perpetuity. We have applied this concept to other ventures in ophthalmology, including corneal refractive surgery and macular degeneration. The key to progress is continued learning from large datasets that can be gleaned from surgeons across the globe.

My colleagues Albert Jun, MD, PhD, and John Ladas, MD, PhD, and I started an AI health care company that began with IOL formulas and IOL calculations and resulted in a U.S. patent recently awarded to Advanced Euclidean Solutions for the unique and novel methodology to improve IOL calculations as well as an in-the-cloud device to seamlessly improve all formulas or approaches to these refractive calculations.

While it may seem obvious now, it was not when we started our work more than a decade ago. Our journey started with an original paper that demonstrated that we could think of formulas in multiple dimensions, but it then rapidly advanced as we realized we could adjust any formula now or in the future. And by any formula, we mean vergence formula, thick or thin lens based, paraxial ray based or even an AI-generated formula.

For IOL calculations, over the past 40 years, surgeons have relied on adjusting A-constants to improve their surgical results. It has been achieved by relying on individual postoperative refractive results or a database of many surgeons (like the online ULIB database). In fact, if a surgeon did not adjust their A-constant and used just what was labeled on the box, it would have been considered substandard care. It is now well known that adjusting an A-constant equally for all types of eyes is inadequate, and individuals and companies have recognized this. Let me show you why this does not make sense now. Four known variables — axial length, corneal power, lens thickness and anterior chamber depth — have an influence. As an aside, there are many other variables that have been used or proposed, and likely many more will come in the future. All of these variables influence the success of the formula and are intimately related. There is no human who could adjust a formula for a specific eye.

While the concept was simple and elegant, the algorithms we initially used were in their early stages of development. Since then, the algorithms to achieve our “adjustments” have become more sophisticated and have kept evolving. In addition, coding of these programs has become easier and more efficient with large language models to help with the process.

Our patent involves AI-based optimization of any formula (including our own) for calculating IOL power in cataract surgery. In addition to working on any formula, the intellectual property is secured for working on as few as two input variables (ie, axial length and corneal power) or an infinite number of variables that exist now or will ever exist (Figure 1). One facet of our methodology is that no matter what formula one starts with, over time all formulas will evolve into a perfect formula that we call the Singularity as more outcomes and variables are introduced.

The allowed patent also included a virtual “in the cloud” device that accumulates data, stores it and uses our methodology to improve itself (Figure 2). These allowed claims are separate but important because this secures the most likely and efficient way for the data and the process to evolve.

While this was our first patent, we have others in the pipeline that will help our field as well as our patients over time. Our intellectual property in corneal refractive surgery is similar to the IOL calculation patent in that it uses outcomes to adjust future treatment plans using deep learning. Our retina projects are a bit different because they help to predict the various treatment outcomes and guide future interventions as new modalities come into existence in a continually self-optimizing algorithm. Think Waze-like directions that are constantly finding the best pathway to get to a particular destination. Or if you prefer this analogy, playing chess with a chess engine by your side.

We are at a point in medicine where physicians can use AI and deep learning as part of what is called “distributed cognition,” the theory that describes how information processing is spread across people, their environment and their tools: We believed 10 years ago and now more than ever that this is a pathway to better outcomes in all fields of medicine, especially ophthalmology.

• For more information:
• Uday Devgan, MD, in private practice at Devgan Eye Surgery and a partner at Specialty Surgical Center in Beverly Hills, California, can be reached at devgan@gmail.com; website: www.CataractCoach.com.

Published by:

Read more about

This post was originally published on this site