International Journal of Drug Delivery Technology · May 29, 2026
Retinal Imaging as a Non-Invasive Biomarker for Alzheimer’s Disease: A Survey of Segmentation and Deep Learning Approaches
By D NR, Sarsamba MC
Retinal imaging has become an essential non-invasive method of diagnosis of ophthalmic and neurodegenerative diseases, such as Alzheimer disease (AD) and mild cognitive impairment (MCI). The recent developments in the field of deep learning have facilitated the automated segmentation of the retinal vessels and optic disc, which benefits the detection of diseases and clinical decisions made. The paper overviews the current state of art in retinal image analysis, including convolutional neural networks (CNNs), transformer - based models, and state-space models including Mamba. Also, it examines how retinal imaging modalities such as fundus photography, optical coherence tomography (OCT), OCT angiography (OCTA), and hyper spectral imaging can be used in early detection of neurodegenerative diseases. The review indicates newer dynamics in the field of hybrid architectures, multi-scale and multi-view feature fusion, and explainable AI techniques, and points out issues concerning the lack of data, quality of features, computational demands, and clinical interpretability. Through a critical evaluation of current approaches and shortcomings, the paper offers a broad vision to inform future studies in automated retinal image recognition and detection of early AD.