Improved Semantic Segmentation of Tuberculosis-consistent findings in Chest X-rays Using Augmented Training of Modality-specific U-Net Models with Weak Localizations

TL;DR

This study enhances TB-related lesion segmentation in chest X-rays by augmenting training data with weak localizations, improving U-Net model performance across multiple datasets for better diagnostic support.

Contribution

It introduces a novel training augmentation using weak localizations from a classifier to improve U-Net segmentation of TB findings in chest X-rays.

Findings

Augmented training improved segmentation accuracy significantly (p < 0.05).

Models generalized well across different datasets.

Modality-specific training outperformed pretraining on natural images.

Abstract

Deep learning (DL) has drawn tremendous attention in object localization and recognition for both natural and medical images. U-Net segmentation models have demonstrated superior performance compared to conventional handcrafted feature-based methods. Medical image modality-specific DL models are better at transferring domain knowledge to a relevant target task than those that are pretrained on stock photography images. This helps improve model adaptation, generalization, and class-specific region of interest (ROI) localization. In this study, we train chest X-ray (CXR) modality-specific U-Nets and other state-of-the-art U-Net models for semantic segmentation of tuberculosis (TB)-consistent findings. Automated segmentation of such manifestations could help radiologists reduce errors and supplement decision-making while improving patient care and productivity. Our approach uses the publicly available TBX11K CXR dataset with weak TB annotations, typically provided as bounding boxes, to train a set of U-Net models. Next, we improve the results by augmenting the training data with weak localizations, post-processed into an ROI mask, from a DL classifier that is trained to classify CXRs as showing normal lungs or suspected TB manifestations. Test data are individually derived from the TBX11K CXR training distribution and other cross-institutional collections including the Shenzhen TB and Montgomery TB CXR datasets. We observe that our augmented training strategy helped the CXR modality-specific U-Net models achieve superior performance with test data derived from the TBX11K CXR training distribution as well as from cross-institutional collections (p < 0.05).