Deep learning retinal imaging model identifies poor brain health among older adults without dementia

• The i-Cog Brain Health is a validated deep learning model for differentiating Alzheimer’s disease dementia from cognitively unimpaired subjects based on retinal photographs. • This study aimed to apply the i-Cog Brain Health on healthy community older subjects and assess whether altered retinal microvasculopathy can be detected in nondemented older adults. • Results from 185 subjects showed that the i-Cog Brain Health reflected older age and wider venular branching width, both of which are associated with the incidence dementia. • The i-Cog Brain Health showed the potential to differentiate retinal features associated with dementia at an early stage and serve as a risk stratification tool. The i-Cog Brain Health is a validated deep learning model for differentiating Alzheimer’s disease dementia from cognitively unimpaired subjects based on retinal photographs. This study aimed to apply the i-Cog Brain Health in subjects without dementia and assess whether this tool may detect alterations in the retinal vessel network in healthy older adults. Community subjects were recruited from the BEAT AD ( B rain Health E valuation A nd T ailor-made Measures for Prevention of A lzheimer’s D isease) service programme. Clinical data, vascular risk factors, lifestyle information and cognitive function were assessed. Tailor-made recommendations were provided to optimise risk factor control. Fundus photographs were obtained using the Topcon NW500 non-mydriatic retinal camera. Subjects were classified into “positive” or “negative” using i-Cog Brain Health based on quantitative measurements of retinal vessels. Among the 185 subjects (mean age: 68.14 ± 5.17 years; males: 32.97%), 29 (15.68%) were classified as “positive” by i-Cog Brain Health. Those subjects were significantly older (p = 0.001) and had wider venular branching width (p = 0.008). Regression analyses showed the venule branching coefficient significantly predicted i-Cog Brain Health “positive” cases (OR 1.54, 95% CI: 1.05-2.27, p = 0.027), after adjustments for age and mean arterial pressure. The i-Cog Brain Health reflected older age and wider venular branching width, which are associated with dementia. The i-Cog Brain Health showed the potential to differentiate retinal features associated with dementia at an early stage and serve as a risk stratification tool.