Balanced-MixUp for Highly Imbalanced Medical Image Classification

TL;DR

This paper introduces Balanced-MixUp, a novel data sampling method that combines instance-based and class-based sampling with MixUp regularization to improve classification performance on highly imbalanced medical image datasets.

Contribution

The paper proposes Balanced-MixUp, a new sampling technique that enhances learning from minority classes in imbalanced datasets by integrating balanced sampling with MixUp.

Findings

Balanced-MixUp outperforms traditional sampling methods.

It improves classification accuracy on imbalanced medical datasets.

The method is effective across different CNN architectures.

Abstract

Highly imbalanced datasets are ubiquitous in medical image classification problems. In such problems, it is often the case that rare classes associated to less prevalent diseases are severely under-represented in labeled databases, typically resulting in poor performance of machine learning algorithms due to overfitting in the learning process. In this paper, we propose a novel mechanism for sampling training data based on the popular MixUp regularization technique, which we refer to as Balanced-MixUp. In short, Balanced-MixUp simultaneously performs regular (i.e., instance-based) and balanced (i.e., class-based) sampling of the training data. The resulting two sets of samples are then mixed-up to create a more balanced training distribution from which a neural network can effectively learn without incurring in heavily under-fitting the minority classes. We experiment with a highly imbalanced dataset of retinal images (55K samples, 5 classes) and a long-tail dataset of gastro-intestinal video frames (10K images, 23 classes), using two CNNs of varying representation capabilities. Experimental results demonstrate that applying Balanced-MixUp outperforms other conventional sampling schemes and loss functions specifically designed to deal with imbalanced data. Code is released at https://github.com/agaldran/balanced_mixup .