Improving Multi-Class OCT Classification under Class Imbalance via Class-Weight Balancing
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Class imbalance is a critical issue in multi-class classification, particularly in medical image analysis, where rare diseases are insufficient. In this study, an improved class weight balancing (CWB) formulation, modified from the baseline method in [1], was proposed and applied to an eight-class imbalanced classification task for Optical Coherence Tomography (OCT) images, to solve the imbalance data issue and enhance the performance. The modified CWB formula, derived from conventional weighting strategies, is applied to various convolutional neural network (CNN) architectures. In addition, basic data augmentation (DA) techniques are employed to address the few-shot problem in rare disease classes. The results show that AOCT-NET combined with the modified CWB and basic DA achieves 78.35% of accuracy and 0.7221 of F1-score, outperforming other lightweight CNN models, and demonstrating improvements over the original CWB method. Furthermore, the CWB method effectively addresses data imbalance, particularly in lightweight CNN models. Future work will focus on further enhancing the recognition of rare diseases and mitigating dataset limitations.
Fields of Study & Subjects
Imbalanced Data Classification Techniques
Retinal Imaging and Analysis
Advanced Neural Network Applications
Publication details
- Journal / source
- International Conference on Consumer Electronics
- Pages
- 1–5
- Publication date
- 3 February 2026
- Language
- EN