UWI Mona student builds AI system to detect diabetic eye disease in low-resource settings

Jamaica is marking World Salt Awareness Week with renewed calls to cut sodium in everyday eating, especially for children, as part of broader efforts to protect heart health.

Health specialists say lowering salt in children’s diets now can help reduce heart disease later, particularly among those who are overweight. Parents are urged to cook with little or no added salt, favour fresh herbs and seasonings, limit salty snacks, and read nutrition labels to avoid high-sodium products. Children under 11 should get no more than 2 grams of salt per day from all food sources.

According to the Jamaica Health and Lifestyle Survey, roughly 12,000 Jamaicans have experienced heart attacks, including about 8,000 men and 4,000 women. Experts link risk to smoking, nutrition, rest, physical activity, and environmental influences. Regular smokers face about two to four times the heart-attack risk of the general population, though benefits can begin within 24 hours of quitting and cardiovascular risk may approach that of non-smokers after 10 to 15 years.

Specialists also warn that excess salt can impair blood-vessel function within about half an hour of ingestion, that prolonged sitting raises cardiovascular risk, that short sleep is linked to higher heart-disease rates, and that intense anger can sharply increase heart-attack risk in the following hours. Jamaicans consume an average of 9 to 12 grams of sodium per day, about twice the World Health Organization’s recommended ceiling of less than 5 grams, with an ideal adult target below 2 grams of sodium daily from all sources. Adults are advised to cap intake at about 3 to 5 grams, or half to one teaspoon, per day.

In parallel, Jamaica is advancing use of artificial intelligence in healthcare. At the University of the West Indies Mona campus, final-year engineering student Ricardo Harrison has built a prototype that uses AI-powered computer vision to detect diabetic retinopathy from retina photographs. The year-long capstone project was proposed by collaborators at the University of Leeds, Gerard Loza and Nikita Greenidge, and supervised by Sasha Gay Wright, head of biomedical engineering at UWI Mona.

Diabetic retinopathy, a diabetes complication affecting the eyes, is a leading cause of blindness among working-age adults and a major source of vision loss worldwide for people aged 30 to 64. Integrating AI into diagnosis is seen as important for early detection and timely treatment; experts say visual impairment can be prevented in more than 90 percent of cases when retinopathy is caught and managed promptly.

Harrison’s system captures a retina image with a 20D lens that magnifies the back of the eye 3.3 times, uses a smartphone for imaging, and runs a model hosted on Google to classify severity levels including mild, moderate, proliferative, or healthy findings. The design targets low-resource hospitals where conventional machines are costly and hard to move. It uses the ResNet 152 version 2 deep-learning architecture and includes patient monitoring for vitals such as heart rate, blood pressure, body temperature, and oxygen level.

Harrison, who trained the model with a Kaggle dataset, said the project was his first involving AI and described the work as demanding but completed with faculty and peer support. He estimates the system is about 78 percent ready for hospital integration and hopes external support and a higher-quality camera could push readiness toward 95 percent, making screening more affordable and accessible across the Caribbean where equipment for retinopathy detection remains limited. He said wider hospital use could also raise awareness among patients who may not know they need screening.