As we end another year and welcome 2019, the technology and ideas of yesterday are becoming a reality. Gordon Moore predicted in 1965 the components of transistors would double every year and later on to every 2 years,1 as demonstrated elegantly below. So now that we have the hardware, do we have the software, and data? The answer is, yes!
The birth of artificial intelligence was in 1956 at a Dartmouth conference, which took areas of mathematics, engineering, psychology, economics and political science to give this "idea" a name and mission.2 Today, with the vast amounts of data collection, healthcare has focused in areas of predictive models to have better outcomes.
Many independent optometrists use analytical programs in their practices, typically for financial and practice management purposes. However, those are description analytics, or commonly known as "dashboards." Descriptive analytics focus on what happened historically by examining raw data from multiple sources; however, this does not answer the "why." Diagnostic analytics are then used to drill down into the raw data and use algorithms to find the drivers of the historical data, which is extremely helpful for understanding particular financial risks of a practice.
But now analytics has become predictive, where data from descriptive and diagnostic models are analyzed further to give information about what is "likely" to happen. But user beware, this type of forecasting is only as good as the quality of data that is collected, along with the similarity of situations. Using different regression models and neural networks, predictive analytics is the basis for artificial intelligence (AI) and deep learning. Those AI programs that suggest actions to take to lower risks are called prescriptive analytics, which requires lots of robust data and sophisticated algorithms.
In eye care, there are many areas where AI is being developed and refined: diabetes, glaucoma, cardiovascular disease and dementia. Currently, there is AI for diabetes on the market, both FDA approved as Class 1 and Class 2 devices. The class of the device is based upon the use and risk to a patient as the higher the number indicates higher regulatory control. There are several different companies, such as Retinalyze, Diagnos, Retina AI and IDx, that have become tools to improve outcomes for patients. Of those listed, IDx is the only FDA approved, Class 2 device due to its 87.2% sensitivity and 90.7% specificity, which does not require an interpretation by a physician.3 This type of clearance is concerning for eye care providers, both optometrists and ophthalmologists, because now health care providers who are not diagnosticians in eye care can use this system to check for diabetic retinopathy.
Similarly, Visulytix developed an AI detect for glaucoma called Pegasus-disc where Harvard University has conducted a study called, "Automated Evaluation of Optic Disc Images for Manifest Glaucoma Detection Using Deep-Learning, Neural Network-Based Program."4 It concluded from the sample size of 400 patients that Pegasus detection of glaucoma was equivalent to 2 glaucoma experts, and maybe more sensitive.4
Verily, a subsidiary of Alphabet Inc. and parent to Google, is using machine learning and neural networks from scans of the retinal blood vessels. They have built a dataset consisting of 300,000 patients, where the AI is learning how to determine blood pressure, patient’s age, and whether they smoke from the images. So far, the algorithm is about 70% sensitive. However, the current SCORE (the European cardiovascular disease risk assessment model) method is only 72% sensitive.5
Lastly, eye movements and pupillary responses are being studied as early biomarkers for dementia. Neurotrack has designed an AI that tracks eye movement speed, direction and patterns to monitor cognitive performance in patients diagnosed with early dementia.6 The cloud-based application is designed for at-home monitoring, using a webcam which takes about 5 minutes per day.7
As optometrists, we have to start embracing these technologies to provide better outcomes for our patients. We need to have tolerance of these technological changes to benefit our patients with better outcomes and lower costs. Defending the status quo will have substantial costs if we do not conform to new standards of care.
Bryan M. Rogoff, OD, MBA, CPHM has a unique background in areas of holistic eye care, business management and healthcare reform. He specializes in LEAN clinical management and operations, technology implementation, healthcare strategy, and strategic partnerships. Currently, he serves as a consultant for for the FDA, Immediate Past-President & Education Chairperson for the Maryland Optometric Association, Federal Keyperson and Meetings Committee Member for the American Optometric Association, reviewer for the Council on Optometric Practitioner Education and is the Founder of Eye-Exec Consulting, LLC. To contact Bryan, visit www.eye-exec.com or email firstname.lastname@example.org. He can also be found on LinkedIn, Facebook, Twitter and Instagram.