NEURODEGENERATIVE DISEASE DETECTION USING RETINAL OCT IMAGES

Abstract

Early diagnosis of neurodegenerative diseases like Alzheimer's and Parkinson's disease is still an important clinical challenge given the subtle first signs and the invasive nature of traditional diagnosis techniques. The current study tackles the issue by utilizing retinal optical coherence tomography (OCT) images as an inexpensive, noninvasive biomarker for neurological health. High-resolution OCT allows the detection of retinal layer changes that are correlated to neurodegeneration. The suggested system utilizes deep learning, namely fine-tuned convolutional neural networks (AlexNet, GoogLeNet, and ResNet-50) by transfer learning, to distinguish between healthy and diseased retinal scans. Noise reduction, contrast enhancers, and normalization preprocessing steps are taken to provide the best input quality. The novelty is the incorporation of superior CNN architectures with transfer learning on retinal OCT records for strong, autonomous prediction of Alzheimer's and Parkinson's disease. The outcomes prove high prediction accuracy, highlighting the promise of autonomous retinal biomarker analysis allowing early diagnosis, timely treatment, and improved patient outcomes.