Remidio Innovative Solutions today announced the publication of a landmark independent evaluation of regulatory-approved diabetic retinopathy (DR) artificial intelligence algorithms in The Lancet Digital Health. Led by Professor Alicja Rudnicka, Professor Adnan Tufail, and the UK National Health Service (NHS) team, the study represents the largest real-world assessment of DR AI to date, analysing over 200,000 screening encounters and approximately 1.2 million retinal images.
The evaluation assessed 32 DR AI algorithms, of which eight participated in the final analysis. Remidio’s CE Class IIa approved DR AI was among the algorithms included in this large-scale real-world study.
Clinical performance and screening relevance
The evaluation demonstrated that Remidio’s AI delivers strong sensitivity for referable and sight-threatening diabetic retinopathy, while maintaining high specificity and predictive value, attributes that are essential for population-scale screening programmes.
For referable DR, which includes moderate non-proliferative DR and more serious disease, Remidio’s AI achieved an overall 87 percent sensitivity. Sensitivity increased to approximately 99 percent for moderate-to-severe non-proliferative DR and around 97 percent for proliferative DR, indicating that sight-threatening disease is highly unlikely to be missed. This supports timely referral and intervention to prevent avoidable vision loss.
The AI also demonstrated high specificity and positive predictive value, reducing false positives and unnecessary referrals. With a negative predictive value of approximately 98 percent, the system provides strong reassurance for negative screens, supporting wider coverage and the safe extension of screening intervals without additional burden on health systems.
Dr R. Kim, Senior Consultant, Aravind Eye Hospital, said, “Large-scale diabetic retinopathy screening demands both diagnostic sensitivity and operational discipline. The findings reported in this evaluation are clinically meaningful because they demonstrate that AI can detect sight-threatening disease reliably, while maintaining specificity levels that are essential to avoid overwhelming referral systems. This balance is critical for sustainable screening programmes, particularly in high-volume public health settings.”
Dr Divya, Chief Medical Officer at Remidio Innovative Solutions, said, “This study evaluates AI performance at real-world screening scale rather than in controlled environments. The results show that Remidio’s AI can reliably identify sight-threatening disease while avoiding unnecessary referrals, which is critical for safe and efficient public health screening.”
Equity, scalability, and health system impact
The study found Remidio’s AI performance to be stable across demographic sub-groups, including age, sex, and ethnicity, supporting equitable deployment across diverse populations.
At health-system level, the findings indicate that Remidio’s AI can reduce reliance on human grading by up to approximately 80 percent, allowing specialist capacity to be focused on patients requiring care. By improving true-positive yield and reducing unnecessary referrals, the AI has the potential to significantly improve the cost-effectiveness of diabetic retinopathy screening pathways.
Designed for real-world screening
The results reinforce Remidio’s long-standing focus on developing AI solutions for real-world screening programmes, particularly in public health and resource-constrained settings. As the global burden of diabetes and diabetic retinopathy continues to rise, AI systems that combine high sensitivity with operational efficiency will be essential to building scalable, equitable, and sustainable eye-care models.
