NEC Health’s AI Algorithm Validated in Landmark Lancet Study for NHS Diabetic Eye Screening

The research, led by City St George’s, University of London and Moorfields Eye Hospital NHS Foundation Trust, addressed the critical need for unbiased validation by testing algorithms on a “level playing field” against the real-world complexities of NHS service delivery. 

A benchmark for fairness 

Historically, the adoption of AI in healthcare has been hampered by concerns over bias, specifically whether algorithms perform equally well across different ethnicities and age groups. 

This study rigorously tested eight algorithms. Evaluating over 1.2 million retinal images from more than 200,000 screening visits, the researchers created a secure, independent environment for the tests. The dataset was notably diverse, reflecting the population of the North East London Diabetic Eye Screening Programme. 

The results demonstrated that these AI systems are safe and perform “equitably across different ethnicities and age groups” for the detection of high-risk diabetic retinopathy. Co-lead Professor Alicja R. Rudnicka described the project as a “transparent, trustworthy, and impartial” evaluation that sets a new standard for AI scrutiny. 

High sensitivity for maximum safety 

NEC Software Solutions, whose OptoMize software supports the majority of the UK’s diabetic eye screening programmes, was one of the eight vendors selected to participate in this rigorous evaluation. 

Identified in the study as Vendor E, NEC’s algorithm demonstrated a profile calibrated for maximum patient safety. The NEC algorithm achieved the highest sensitivity (detection rate) of all systems tested for referable diabetic retinopathy, at 98.7%. 

Most critically, the study confirmed that the NEC algorithm detected 100% of proliferative diabetic retinopathy cases (R3M0/M1), the most advanced and dangerous stage of the disease. 

Balancing Safety and Efficiency 

While some AI models prioritise processing speed or workload reduction, NEC’s technology is calibrated to ensure that no potential case is overlooked. The study showed that while the NEC algorithm flagged a higher proportion of images for human review than some competitors, it provided the robust assurance of a 99.4% negative predictive value. This means that when the system clears a patient, clinicians can be exceptionally confident they are disease-free. 

Andy Combes, Development Manager for Health at NEC Software Solutions, commented: “We welcome this independent validation from The Lancet Digital Health. It confirms what we have always believed: that AI must be fair, robust, and above all, safe. By achieving a 100% detection rate for the most severe disease categories, our algorithm acts as the ultimate safety net. It supports clinicians by automating the clear-cut cases, allowing them to dedicate their expertise to the patients who need it most.” 

The path to safe and sustainable AI adoption 

The publication of this landmark study is a crucial first step, confirming the safety and fairness of AI in diabetic eye screening. The next phase, as outlined by the study’s authors, is translating these findings into safe and practical use across the NHS. 

Writing for Digital Health, Professor Rudnicka highlights that for AI to move from policy papers to real NHS pathways, it requires more than technical accuracy; it demands meaningful digital readiness. 

As the provider of OptoMize, the integrated software platform that manages data for the majority of the UK’s diabetic eye screening programmes, NEC Software Solutions already possesses the secure digital infrastructure essential for the successful scaling of AI. This, combined with NEC’s long-standing role in the NHS, means we are uniquely positioned with the capability and established relationships to support the introduction of these algorithms safely and effectively. 

Jenny Witts, Operations Director for Health at NEC Software Solutions, concluded: “Our priority is to support the NHS in its journey toward safe and sustainable AI adoption. This independent evaluation, combined with our proven infrastructure through OptoMize, reinforces the principle that successful adoption relies on supporting rather than replacing the workforce. AI is a powerful tool to free up clinical capacity and standardise repetitive tasks, allowing clinicians to focus on complex patient care and ensuring that no patient is ever left behind.”