Improvement Strategies for Few-Shot Learning in OCT Image Classification of Rare Retinal Diseases

TL;DR

The paper addresses the challenge of data scarcity in few-shot learning for OCT image classification of rare retinal diseases and identifies limitations of CycleGAN-based augmentation. It replaces CycleGAN with U-GAT-IT, introduces data-balancing, and incorporates attention mechanisms (SE and CBAM) into a fine-tuned InceptionV3 backbone to strengthen feature localization in diseased retinal regions. Key contributions include demonstrating that U-GAT-IT with balancing substantially improves balanced accuracy and overall performance, with CBAM delivering the best overall accuracy (approximately 97.85%) and robust performance across both common and rare classes. The proposed framework offers a practical path toward reliable, balanced OCT-based diagnosis in clinical settings, particularly for rare retinal diseases with limited data.

Abstract

This paper focuses on using few-shot learning to improve the accuracy of classifying OCT diagnosis images with major and rare classes. We used the GAN-based augmentation strategy as a baseline and introduced several novel methods to further enhance our model. The proposed strategy contains U-GAT-IT for improving the generative part and uses the data balance technique to narrow down the skew of accuracy between all categories. The best model obtained was built with CBAM attention mechanism and fine-tuned InceptionV3, and achieved an overall accuracy of 97.85%, representing a significant improvement over the original baseline.

Figures (1)

• Figure 1: Feature space visualized using the 3D t-SNE technique. (a) t-SNE visualization of GAN-based augmentation by CycleGAN. (b) t-SNE visualization of GAN-based augmentation by U-GAT-IT.