Lightweight Convolutional Neural Networks for Retinal Disease Classification

TL;DR

This study evaluates lightweight CNN architectures, MobileNetV2 and NASNetMobile, for classifying retinal diseases from fundus images, achieving high accuracy and demonstrating potential for AI-assisted diagnosis.

Contribution

It introduces the application of MobileNetV2 and NASNetMobile with transfer learning for retinal disease classification, emphasizing efficiency and accuracy in a limited dataset context.

Findings

MobileNetV2 achieved 90.8% accuracy

Transfer learning improved model performance

Lightweight CNNs are effective for retinal disease detection

Abstract

Retinal diseases such as Diabetic Retinopathy (DR) and Macular Hole (MH) significantly impact vision and affect millions worldwide. Early detection is crucial, as DR, a complication of diabetes, damages retinal blood vessels, potentially leading to blindness, while MH disrupts central vision, affecting tasks like reading and facial recognition. This paper employed two lightweight and efficient Convolution Neural Network architectures, MobileNet and NASNetMobile, for the classification of Normal, DR, and MH retinal images. The models were trained on the RFMiD dataset, consisting of 3,200 fundus images, after undergoing preprocessing steps such as resizing, normalization, and augmentation. To address data scarcity, this study leveraged transfer learning and data augmentation techniques, enhancing model generalization and performance. The experimental results demonstrate that MobileNetV2 achieved the highest accuracy of 90.8%, outperforming NASNetMobile, which achieved 89.5% accuracy. These findings highlight the effectiveness of CNNs in retinal disease classification, providing a foundation for AI-assisted ophthalmic diagnosis and early intervention.