Predict Cognitive Disorders From Retinal Fundus Images Using Automated Retinal Vasculature Analysis Program

Cognitive disorders are prevalent global health issues. Contemporary diagnosis and biomarkers for cognitive disorders, especially Alzheimer’s disease (AD), involves the detection of Aβ via cerebrospinal fluid or positron emission tomography that are invasive and expensive, which make them unsuitable for disease screening. Features in retinal vasculatures from retinal fundus images are considered to be correlated to brain cognitive status. In this study, based on digital image processing methods, we developed automated program that calculate retinal vasculature parameters from fundus images. Using this approach, we calculated parameters of 136 fundus images from normal control (NC), patients with mild cognitive impairment (MCI), and AD. Multiplicative models were adopted to discriminate MCI & AD patients from NC based on vasculature discrepancies. Using 40-fold cross-validation, our models have an average area under the receiver operating characteristic curve values 0.744. Our program facilitates the analysis of retinal vasculature, from which instructive correlations were revealed between retinal vasculature and cognitive status. Discriminative models for MCI & AD patients consolidate the role of retinal fundus image as a potential biomarker for cognitive disorders.