Multi-label Thoracic Disease Image Classification with Cross-Attention Networks

TL;DR

This paper introduces Cross-Attention Networks (CAN) for multi-label thoracic disease classification in chest X-ray images, effectively handling class imbalance and leveraging only image-level annotations to improve diagnostic accuracy.

Contribution

The paper proposes a novel Cross-Attention Network architecture and a new loss function to enhance multi-label thoracic disease classification from chest X-rays, addressing class imbalance and data scarcity issues.

Findings

Achieves state-of-the-art classification performance.

Effectively handles class imbalance and easy-dominated samples.

Utilizes only image-level annotations for training.

Abstract

Automated disease classification of radiology images has been emerging as a promising technique to support clinical diagnosis and treatment planning. Unlike generic image classification tasks, a real-world radiology image classification task is significantly more challenging as it is far more expensive to collect the training data where the labeled data is in nature multi-label; and more seriously samples from easy classes often dominate; training data is highly class-imbalanced problem exists in practice as well. To overcome these challenges, in this paper, we propose a novel scheme of Cross-Attention Networks (CAN) for automated thoracic disease classification from chest x-ray images, which can effectively excavate more meaningful representation from data to boost the performance through cross-attention by only image-level annotations. We also design a new loss function that beyond cross-entropy loss to help cross-attention process and is able to overcome the imbalance between classes and easy-dominated samples within each class. The proposed method achieves state-of-the-art results.