From Darkness to Light: How AI is Redefining Glaucoma Screening for All
By Ziyue Guan, Serene Zhang, and Wenyu Seet, GRC 2024 Global Essay Competition Top 5
The world is still there, but it is slipping away piece by piece. I see the sun but not the horizon, the face of my child but not her outstretched hand. The colors of the world grow muted like watercolor bleeding in the rain, beginning with a blur in my vision—a creeping fog. It will go away. But the fog thickens.
And yet, there’s light—not hope, necessarily, but light. I want to keep moving, to keep reading, to keep seeing. I want my world to stay open.
This is the quiet reality of glaucoma—an affliction that 111.8 million people will soon experience in 2040. Known as the 'silent thief of the night,' glaucoma increases intraocular pressure (IOP) due to a buildup of fluid in the eye’s aqueous humor, stealthily damaging the optic nerve and becoming the second leading cause of blindness worldwide. Alarmingly, nearly 50% of those affected remain unaware of their condition. In its early stages, glaucoma is asymptomatic. The effects of glaucoma creep in subtly, with the loss of vision typically beginning in peripheral vision. Without the intervention of successful treatment, this can progress to total vision loss.
Yet today, there is no proven way of recovering vision loss caused by glaucoma. The urgency is stark: act now or risk losing the beauty, the light, the life that sight allows.
But, this is easier said than done. Glaucoma remains a silent thief of sight, yet for those in developing countries, the silence is deafening. In these regions, 90% of glaucoma cases go undetected because for far too many, access to eye care is a privilege, not a right. This inequity manifests in dire consequences: in rural Sub-Saharan Africa, blindness is four times more prevalent than in wealthier countries. Why? A single glaucoma screening can cost over $100—a prohibitive price for families earning just $2-$3 per day. For families living in rural communities, a parent must choose: to sacrifice a month’s wages for a distant hope of saving their sight, or let it be? The lack of affordable services—if such care is even available, leaves them with no choice but to let darkness encroach.
Compounding the problem is a global shortage of ophthalmologists. Based on a study from the HRSA’s National Center, the total ophthalmology supply is projected to decrease by 12% of full-time equivalent (FTE) ophthalmologists, while the total demand is projected to increase by 24% of FTE ophthalmologists from 2020 to 2035, representing a supply and demand mismatch of 30% workforce inadequacy. With our aging society, the prevalence of age-related eye conditions like glaucoma is on the rise, intensifying the demand for ophthalmic services. Moreover, the uneven distribution of ophthalmologists, with a concentration in urban areas, forces millions in rural regions to face the devastating impacts of untreated eye diseases.
To address these challenges, we must turn to innovation. Today, Artificial Intelligence (AI) is transforming glaucoma detection by leveraging advanced machine learning techniques, particularly deep convolutional neural networks (CNNs), to identify subtle signs of the disease that go unnoticed by the human eye. Machine learning has proven to be an incredibly potent
technique in image classification over the past decade. In particular, ever since AlexNet was introduced in 2012, deep CNNs have become the norm for image recognition tasks.
How do CNNs work? Intuitively, CNNs are modeled after how humans perceive images. These models apply multiple convolutional filters across the input image, learning local features (e.g., texture, edges) at early layers and progressively capturing more complex, high-level representations at deeper layers. In glaucoma detection, these learned representations can reveal subtle structural changes in the optic nerve head or nerve fiber layer that might be imperceptible to the human eye.
Thus, we designed a 3D prototype, GoLens, that utilizes an end-to-end machine-learning pipeline to detect glaucoma with high precision and accessibility. By integrating CNNs, which mimic human visual perception, into smartphone-based imaging systems, AI can enable efficient analysis of retinal images, detecting subtle structural changes in the optic nerve head or nerve fiber layer.
Retinal images collected from clinical databases are preprocessed to reduce noise and artifacts (ie. speckles), hence normalizing pixel intensity ranges that make them suitable for consistent analysis.
Popular CNN architectures like ResNet can then identify and extract fundamental features like texture and edges in retinal layers and progress to capture high-level structural information at deeper layers. By training on expertly labeled datasets (the model’s prediction is measured by a simple zero-one loss function during each iteration) and refining through validation, AI provides remarkable accuracy in classifying images as either “healthy” or “glaucomatous.”
Flowchart of GoLens AI integration
AI also tackles the two key issues we identified in glaucoma care.
First, the global ophthalmologist shortage alongside rising glaucoma cases has created a critical eye care gap. To address this, AI can triage patients based on the urgency and severity of their condition, in order to prioritize those who require immediate medical intervention. This is especially crucial in regions where there is a much greater need for ophthalmic care than there are trained specialists.
Second, the disproportionate access to eye care. AI can provide high-quality eye screenings to rural populations, bringing AI-powered diagnostic tools such as GoLens to the doorstep of those who once lived two days away from the nearest clinic. Early diagnosis can make a life-changing difference.
Simultaneously, AI prioritizes affordability for these households, who in the past, had to forgo vital eye care altogether. For a fraction of the price of one conventional eye exam, an entire household can access regular glaucoma screenings using GoLens. This ensures that more
people can detect eye conditions early and transform eye care from an unattainable luxury into an accessible necessity.
And now, I see the world differently. In the details, I can grasp the creases of my daughter’s smile, the way sunlight pools on her hand.
This is a fight for dignity, equality and the basic human right to see.
Bibliography
He, K., Zhang, X., Ren, S., & Sun, J. (2015, December 10). Deep Residual Learning for Image Recognition. arXiv.org. https://arxiv.org/abs/1512.03385
Krizhevsky, A., Sutskever, I., Hinton, G. E., & University of Toronto. (n.d.). ImageNet Classification with Deep Convolutional Neural Networks. ImageNet Classification With Deep Convolutional Neural Networks https://proceedings.neurips.cc/paper_files/paper/2012/file/c399862d3b9d6b76c8436e924a68c45 b-Paper.pdf
National Eye Institute. (2024, December 16). Glaucoma | National Eye Institute. https://www.nei.nih.gov/learn-about-eye-health/eye-conditions-and-diseases/glaucoma#:~:text=What%20is%20glaucoma%3F,a%20comprehensive%20dilated%20eye%20exam
kwest@preventblindness.org. (2024, October 17). New Study Finds Higher Prevalence of Glaucoma Than Previously Estimated - Prevent Blindness. Prevent Blindness. https://preventblindness.org/new-glaucoma-prevalence-study/
Jackson, A. (2024, November 8). Glaucoma Facts And Stats - Glaucoma Research Foundation. Glaucoma Research Foundation. https://glaucoma.org/articles/glaucoma-facts-and-stats World Glaucoma Week » What is glaucoma? (n.d.). https://www.worldglaucomaweek.org/what-is-glaucoma/
Tham, Y., Li, X., Wong, T. Y., Quigley, H. A., Aung, T., & Cheng, C. (2014). Global Prevalence of Glaucoma and Projections of Glaucoma Burden through 2040. Ophthalmology, 121(11), 2081–2090. https://doi.org/10.1016/j.ophtha.2014.05.013