Seeing The Future: Artificial Intelligence For Age-Related Macular Degeneration
(Posted on Saturday, July 22, 2023)
Originally published on Forbes on 7/17/2023
This story is part of a series on the current progression in Regenerative Medicine. This piece is part of a series dedicated to the eye and improvements in restoring vision. More specifically, this piece continues our discussion of age-related macular degeneration.
In 1999, I defined regenerative medicine as the collection of interventions that restore to normal function tissues and organs that have been damaged by disease, injured by trauma, or worn by time. I include a full spectrum of chemical, gene, and protein-based medicines, cell-based therapies, and biomechanical interventions that achieve that goal.
Identifying those at high risk of age-related macular degeneration is challenging due to a shortage of specialists and the need for clinical expertise to evaluate retinal imaging. But emerging technologies like artificial intelligence/machine learning (AI/ML) pave the way for more accurate and efficient screening methods.
These AI systems can autonomously evaluate one’s risk for developing late age-related macular degeneration and help with prompt diagnosis and treatment. As this technology advances, early intervention can be a game-changer for patients at high risk of disease, helping them maintain their vision and quality of life.
What is Macular Degeneration?
Macular degeneration is a chronic eye condition that affects the macula, the central part of the retina responsible for central vision. This disease is generally age-related, occurring more frequently in individuals over 50. Several risk factors, including genetics, smoking, high blood pressure, obesity, lack of exercise, and poor diet, have been associated with an increased risk of developing macular degeneration.
Over 200 million people worldwide have been diagnosed with this condition, making it one of the leading causes of vision loss in adults. Some early signs and symptoms of macular degeneration include blurred vision, difficulty reading, seeing straight lines as distorted, and experiencing dark patches in the central visual field. Those at risk of developing macular degeneration should prioritize regular check-ups with their eye doctor to receive the appropriate diagnosis and treatment plan to manage the disease effectively.
Identifying the disease in its early stages can facilitate timely intervention and management, preventing irreversible vision loss. With recent advances in artificial intelligence (AI) and machine learning, healthcare providers can now effectively screen patients for age-related macular degeneration.
Current AI Innovations for Retinal Disorders
iHealthScreen develops AI and machine learning-based diagnostic systems for identifying age-related macular degeneration. Their flagship product, iPredict™, enables primary care providers to screen patients for age-related macular degeneration and predict which individuals with early disease are more likely to experience vision loss.
The system was developed using over 93,000 color fundus photos from the landmark Age-related Eye Disease Study (AREDS) and trained using deep-learning algorithms. The resulting model was validated and tested using an additional 23,495 images from AREDS.
To use iPredict™, primary physicians use a fully automated fundus camera to capture color images of a patient’s retinas. These images are then securely sent to a centralized server and analyzed using iPredict™. Based on the analysis, a report classifies the patient as either referable or non-referable for age-related macular degeneration. The system assigns a prediction score for referable patients, quantifying their risk of developing the late-stage disease within the next year or two.
In clinical trials, iPredict™ predicted a 2-year risk for progression to late age-related macular degeneration with 86% and 84% accuracy. The system’s accuracy in a non-specialist setting is being evaluated in four primary care clinics in the New York City area. iPredict’s™ screening model has been prospectively validated and submitted to the FDA for clearance to market the system to primary care practices by the end of 2023.
In collaboration with National Center for Biotechnology Information colleagues, NEI researchers developed another AI-based system for predicting late age-related macular degeneration. Both iHealthScreen and the NIH system are trained to look for reticular pseudodrusen, a type of lesion that causes a spotted pattern in the macula and is associated with a higher risk for progression to late disease.
An Analysis of Artificial Intelligence in Identifying Age-Related Macular Degeneration
Early detection and treatment of age-related macular degeneration are crucial for preventing irreversible vision impairment, and AI-based tools can aid in disease management. A recent analysis focused on using AI-based algorithms for detecting the disease in fundus images. The study found that the algorithms were almost as practical as retinal specialists in detecting the disease.
The AREDS database – with 130,000+ fundus photographs – is frequently used in AI studies of age-related macular degeneration identification. While widely appreciated by researchers, it’s noteworthy that the database was established in the 80s and lacks the nuanced understanding of hard drusen and age-related changes in the current clinical classification of the illness. Therefore, some obsolete images may compromise their suitability for large-scale AI models. Furthermore, note that some of the photographs were digitized from film.
The study discovered that the diagnostic prowess of AI needs to be improved in studies that employ larger validation datasets. It’s crucial to note that said models were evaluated only on research data sets. AI-based screening in larger populations differs significantly under distinct settings and conditions, necessitating a cautious approach. The performance of AI-based algorithms varied considerably in the studies included owing to various factors, including algorithm architecture, data size for training and validation, image quality, and lack of reference standards for defining age-related macular degeneration.
Limitations of Using Artificial Intelligence for Age-Related Macular Degeneration Diagnosis and Predictions
Artificial Intelligence has the potential to detect age-related macular degeneration from multimodality imaging data, including fundus photographs, spectral domain optical coherence tomography (SD-OCT), and angio-OCT. Integrating AI-based software into a fundus camera can help ophthalmologists reduce their workload, lower the likelihood of misdiagnoses, and detect early-stage macular degeneration more efficiently, especially in remote areas lacking skilled specialists. Nonetheless, AI algorithms are typically developed to detect only one disease or sign, trained on limited data sets, and their effectiveness relies on the quality of the image for accurate diagnosis.
There are additional challenges to implementing AI software, such as the feasibility and performance of the software compared to clinical physicians, patient trust in machines, and the potential issues of a “black box” system, where physicians might miss false-negative cases, and patients must understand that referrals are necessary if age-related macular degeneration symptoms develop. Moreover, these techniques are not yet suitable for primary care screening due to their higher cost than non-mydriatic automatic cameras, nor are they helpful for retinal specialists who can read the images themselves.
Nevertheless, research in this field holds great promise for the early diagnosis and treatment of age-related macular degeneration, which can improve patient outcomes and even save lives.
To learn more about the eye, read more stories at www.williamhaseltine.com