Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LTCXNet: Advancing Chest X-Ray Analysis with Solutions for Long-Tailed Multi-Label Classification and Fairness Challenges

Chin-Wei Huang, Mu-Yi Shen, Kuan-Chang Shih, Shih-Chih Lin, Chi-Yu Chen, Po-Chih Kuo

TL;DR

It is demonstrated that LTCXNet improves the performance of CXR interpretation across all classes, especially enhancing detection in rarer classes like `Pneumoperitoneum' and `Pneumomediastinum' by 79\% and 48\%, respectively.

Abstract

Chest X-rays (CXRs) often display various diseases with disparate class frequencies, leading to a long-tailed, multi-label data distribution. In response to this challenge, we explore the Pruned MIMIC-CXR-LT dataset, a curated collection derived from the MIMIC-CXR dataset, specifically designed to represent a long-tailed and multi-label data scenario. We introduce LTCXNet, a novel framework that integrates the ConvNeXt model, ML-Decoder, and strategic data augmentation, further enhanced by an ensemble approach. We demonstrate that LTCXNet improves the performance of CXR interpretation across all classes, especially enhancing detection in rarer classes like `Pneumoperitoneum' and `Pneumomediastinum' by 79\% and 48\%, respectively. Beyond performance metrics, our research extends into evaluating fairness, highlighting that some methods, while improving model accuracy, could inadvertently affect fairness across different demographic groups negatively. This work contributes to advancing the understanding and management of long-tailed, multi-label data distributions in medical imaging, paving the way for more equitable and effective diagnostic tools.

LTCXNet: Advancing Chest X-Ray Analysis with Solutions for Long-Tailed Multi-Label Classification and Fairness Challenges

TL;DR

It is demonstrated that LTCXNet improves the performance of CXR interpretation across all classes, especially enhancing detection in rarer classes like `Pneumoperitoneum' and `Pneumomediastinum' by 79\% and 48\%, respectively.

Abstract

Chest X-rays (CXRs) often display various diseases with disparate class frequencies, leading to a long-tailed, multi-label data distribution. In response to this challenge, we explore the Pruned MIMIC-CXR-LT dataset, a curated collection derived from the MIMIC-CXR dataset, specifically designed to represent a long-tailed and multi-label data scenario. We introduce LTCXNet, a novel framework that integrates the ConvNeXt model, ML-Decoder, and strategic data augmentation, further enhanced by an ensemble approach. We demonstrate that LTCXNet improves the performance of CXR interpretation across all classes, especially enhancing detection in rarer classes like `Pneumoperitoneum' and `Pneumomediastinum' by 79\% and 48\%, respectively. Beyond performance metrics, our research extends into evaluating fairness, highlighting that some methods, while improving model accuracy, could inadvertently affect fairness across different demographic groups negatively. This work contributes to advancing the understanding and management of long-tailed, multi-label data distributions in medical imaging, paving the way for more equitable and effective diagnostic tools.

Paper Structure

This paper contains 22 sections, 2 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Overview
  4. Dataset
  5. ConvNeXt and ML-Decoder
  6. Augmentation
  7. Ensemble Learning
  8. Evaluation Metrics
  9. Implementation Details
  10. Results and Discussion
  11. Performance Evaluation
  12. Backbone Evaluation
  13. Ablation Study
  14. Grad-CAM Visualization
  15. Fairness Evaluation
  16. ...and 7 more sections

Figures (5)

  • Figure 1: Overview of the proposed method: input image processed by three branches (Head, Tail, All), followed by individual predictions and final ensembled prediction (the length of prediction block indicates the output class number).
  • Figure 2: Class distribution in the Pruned MIMIC-CXR-LT dataset.
  • Figure 3: ROC Curve and AUC score for 19 classes separated by Head and Tail classes.
  • Figure 4: Comparison of LTCXNet and baseline (ConvNeXt): testing AP across disease conditions in CXRs sorted by frequency, highlighting top 3 improvement classes.
  • Figure 5: Grad-CAM visualization of (a) Pleural Effusion, (b) Pneumonia, and (c) Cardiomegaly.