FEW DEVELOPMENTS in eye care are as exciting as the potential impact of artificial intelligence (AI) on the future of specialty contact lenses. AI’s capacity for prediction and modeling makes specialty lenses perfect for using large models to maximize patient and practitioner success.
AI is very good at looking over many data points, then computing, predicting, and solving problems based on prior knowledge or outcomes. This is precisely what specialty lens fitters do. Eyecare practitioners (ECPs) determine the patient’s problem, review topography and eye health, then fit the patient using diagnostic lenses, similarly computing and predicting. Finally, they order a lens based on their experience and prior knowledge of how well lenses have worked for previous patients.
In specialty lens design, accurate topographies and shape factors are helpful for fitting lenses. This is already being done. Many topographers allow an ECP to use a module to develop the best lens fit. The computer uses intelligence to match the programmed fit of the lens based on parameters from the topography (or other software). While this method is great, its effectiveness can be limited by measurement errors, errors on the part of the programmer, and/or the ECP, who may “override” the system.
When AI is part of the equation, practitioners will likely send the data points of multiple topographies to a central database. The AI system can analyze the topography (or other imaging) for errors or false data points. It can then explore an extensive database of different data points (thousands to millions) to calculate the curvature and complete lens system that will maximize the success of the lens on the eye. After reviewing the data, the ECP will likely come up with a diameter based on corneal and aperture sizes. The system can review similar eyes and situations (other data points that may have been entered, like osmolarity tear film measurements) to designate a successful material and define other parameters.
The system can then track the success of the fit based on data points the ECP enters about the patient upon dispensing. The system will learn from remake rates, comfort concerns, and visual changes to more accurately predict and model better fits for patients when it designs the next set of lenses.
As with anything significant, there will always be obstacles. One major hurdle is the creation of the database. While contact lens laboratories have databases of patients (and maybe their topographies), privacy concerns can come into play with the use of this data. For this reason, ECPs will likely see the pace of innovation in specialty lenses with AI differ from one country to another.
When we can learn from millions of patients rather than the last hundred that we saw, the success of our practices and patients will improve, and AI can get us there.
