Automatic retinal imaging had good overall performance in detecting age-related white matter changes (ARWMC) and localization of white matter hyperintensities (WMH). The machine learning approach is a convenient and non-invasive tool with potential use in screening and early detection of ARWMC.
Why this matters
WMH detected via brain magnetic resonance imaging (MRI) are an increasingly well-known manifestation of cerebral small vessel disease. The presence and growth of WMH over time (age-related white matter changes [ARWMC]) is associated with increased risk of dementia and stroke in later years. Importantly, WMH are reversible, meaning dementia could potentially be prevented in people where ARWMC are detected early, and vascular risk factors managed appropriately.
Brain MRI is not appropriate as a screening tool for early detection of WMH. Characteristics of retinal blood vessels are known to correlate with the presence of cerebral small vessel disease; it is possible that analysis of retinal images using machine learning approaches could present a viable alternative to brain MRI.