Multi-loss ensemble deep learning for chest X-ray classification

TL;DR

This study benchmarks various loss functions for multi-class chest X-ray classification, proposes improved loss functions, and demonstrates that ensemble methods significantly enhance diagnostic accuracy and interpretability.

Contribution

It introduces novel ensemble strategies and evaluates loss functions specifically tailored for multi-abnormality chest X-ray classification, improving upon existing methods.

Findings

Weighted model ensembles outperform individual models.

Ensembles achieve MCC of 0.9068, surpassing state-of-the-art.

Localization confirms models learn meaningful disease features.

Abstract

Medical images commonly exhibit multiple abnormalities. Predicting them requires multi-class classifiers whose training and desired reliable performance can be affected by a combination of factors, such as, dataset size, data source, distribution, and the loss function used to train the deep neural networks. Currently, the cross-entropy loss remains the de-facto loss function for training deep learning classifiers. This loss function, however, asserts equal learning from all classes, leading to a bias toward the majority class. In this work, we benchmark various state-of-the-art loss functions that are suitable for multi-class classification, critically analyze model performance, and propose improved loss functions. We select a pediatric chest X-ray (CXR) dataset that includes images with no abnormality (normal), and those exhibiting manifestations consistent with bacterial and viral pneumonia. We construct prediction-level and model-level ensembles, respectively, to improve classification performance. Our results show that compared to the individual models and the state-of-the-art literature, the weighted averaging of the predictions for top-3 and top-5 model-level ensembles delivered significantly superior classification performance (p < 0.05) in terms of MCC (0.9068, 95% confidence interval (0.8839, 0.9297)) metric. Finally, we performed localization studies to interpret model behaviors to visualize and confirm that the individual models and ensembles learned meaningful features and highlighted disease manifestations.