tsta

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Diagnostic and Classification Analysis of Retinal Diseases Using OCT Imaging: Focus on Diabetic Retinopathy and Overlap with Other Retinal Disorders

Authors :

Fatemeh Reyhani¹ Yashar Amizadeh² Ata Jodeiri³

1- دانشگاه علوم پزشکی تبریز 2- بیمارستان مهر تبریز 3- دانشگاه علوم پزشکی تبریز

Keywords :

Optical coherence tomography (OCT)،diabetic retinopathy (DR)،deep learning،Vision Transformer (ViT)،data imbalance،disease overlap،early detection

Abstract :

Diabetic retinopathy (DR) is one of the leading causes of blindness worldwide, and early detection is critical to preventing disease progression and reducing vision loss. However, diagnosing DR accurately is challenging due to its overlap with other retinal conditions and the imbalance in class distributions. This study uses optical coherence tomography (OCT) images to distinguish between normal retina (NRM) and DR. We evaluated several deep learning models, including Vision Transformer (ViT), DenseNet121, ResNet50, and EfficientNet, applying techniques like Mixup and Focal Loss to address data imbalance and overlapping diseases. The results show that the ViT model achieved a validation accuracy of 90.80% ± 0.93% in scenarios without overlapping diseases and 90.75% ± 1.74% in cases with disease overlap. ViT demonstrated a remarkable ability to detect subtle DR features in images that appeared normal features that even human experts often struggle to identify. These findings highlight the potential of ViT for early diabetic retinopathy detection and its promising application in patient screening and clinical management.