Ocular Disease Classification Using CNN with Deep Convolutional Generative Adversarial Network
Arun Kunwar, Dibakar Raj Pant, Jukka Heikkonen, Rajeev Kanth
TL;DR
The study tackles limited and imbalanced fundus image datasets for ocular disease classification by employing DCGAN-based data augmentation to synthesize images for myopia, glaucoma, and cataract. A CNN classifier, VGG-16, is trained on the augmented data and evaluated on real images, achieving an overall accuracy of $84.6\%$ with per-class accuracies of $78.6\%$, $88.6\%$, and $84.6\%$ respectively. The results suggest GAN-based augmentation improves generalization in data-scarce medical imaging tasks, though limitations include GAN-generated image noise and 64×64 resolution; future work proposes denoising autoencoders to further boost performance. This work demonstrates a practical pathway to leverage synthetic data for more reliable ocular disease classification in settings with restricted real data.
Abstract
The Convolutional Neural Network (CNN) has shown impressive performance in image classification because of its strong learning capabilities. However, it demands a substantial and balanced dataset for effective training. Otherwise, networks frequently exhibit over fitting and struggle to generalize to new examples. Publicly available dataset of fundus images of ocular disease is insufficient to train any classification model to achieve satisfactory accuracy. So, we propose Generative Adversarial Network(GAN) based data generation technique to synthesize dataset for training CNN based classification model and later use original disease containing ocular images to test the model. During testing the model classification accuracy with the original ocular image, the model achieves an accuracy rate of 78.6% for myopia, 88.6% for glaucoma, and 84.6% for cataract, with an overall classification accuracy of 84.6%.