HEALTH AND TECHNOLOGY
AI and smartphones help prevent blindness
Read Time: 2 minutes
The Barcelona Supercomputing Center is developing an algorithm that will help to make eye exams that detect visual impairments quickly, early and easily.
One in 20 people worldwide lives with some sort of visual impairment that hinders their ability to experience the world in all its multicolor splendor. Yet researchers say nearly 80 percent of visual impairment cases can be avoided through preventative measures.
This discrepancy can be traced to longstanding diagnostic practices that sometimes fail to detect early-stage retinal diseases. This issue is further complicated as patients from the countries most riddled with cases of visual impairment often don’t have access to regular vision screenings due to a lack of access to medical infrastructure.
And this situation stands to worsen. Complications from diabetes are one of the leading causes of visual impairment, and the number of those impacted by the disease is expected to double between 2000 and 2030.
Researchers at the Barcelona Supercomputing Center (BSC) have set out to tackle this diagnostic issue by assisting ophthalmologists in early-stage diagnostics by using artificial intelligence (AI) to give researchers insight into sight itself.
BSC researchers developed a machine-learning model that stands to improve the speed, accuracy, and availability of vision diagnostics. This innovation could one day enable patients to self-administer a vision exam using a smartphone, pre-screening for a series of pathologies like diabetic retinopathy, glaucoma, and macular degeneration. Patients can then take this preliminary information to an ophthalmologist who can administer the care they require.
It’s difficult to train an AI neural network to detect certain diseases when datasets for those pathologies are limited. To get around a lack of “clean data” for some visual impairments, BSC used a Transfer Learning application jointly developed by BSC and Lenovo at the Lenovo AI Innovation Center in Morrisville, North Carolina (USA).
Developed on Lenovo systems, the AI triage application allows the network to store knowledge, acquired while solving one problem, and then apply it to a different, but related issue with limited data availability. For example, data gained by the neural network while learning to recognize daisies could then be applied to roses. Transfer learning effectively reduces the algorithm training time, research hours, and the costs associated with development (supercomputing requires immense amounts of energy and capital).
BSC researchers developed a machine-learning model that stands to improve the speed, accuracy, and availability of vision diagnostics.
The BSC algorithm will improve as the system becomes more widely available; the more eye scans input into the network, the greater the algorithm’s diagnostic capabilities. The technology has the power to transform other industries, as well, says Dario Garcia-Gasulla, a postdoctoral researcher at the Barcelona Supercomputing Center.
“Scaling the design, training, and validation of machine learning models to study these vision issues may be challenging,” he said. “But the impact of potential solutions for this is huge, as the same challenge is found within other medical domains and many other industrial applications.”
Self-diagnostics is just the first step in tackling these complex medical issues. Armed with this data, physicians and researchers will have access to an immensity of data that will improve outcomes for future generations and empower patients to take ownership over their health.
Explore Lenovo
Technology
Lenovo Artificial Intelligence
Lenovo High Performance Computing
Lenovo PCs & Tablets
Additional Stories
