X-ray Dissectography Improves Lung Nodule Detection

TL;DR

This paper introduces X-ray dissectography, a digital technique that enhances lung nodule detection in radiographs by isolating lung structures, leading to a significant increase in detection accuracy.

Contribution

The study presents a novel X-ray dissectography method combined with a collaborative detection network to improve lung nodule detection in 2D and 3D space.

Findings

20+% increase in average precision over baseline

Effective suppression of irrelevant structures in X-ray images

Potential to improve chest radiograph diagnostic workflows

Abstract

Although radiographs are the most frequently used worldwide due to their cost-effectiveness and widespread accessibility, the structural superposition along the x-ray paths often renders suspicious or concerning lung nodules difficult to detect. In this study, we apply "X-ray dissectography" to dissect lungs digitally from a few radiographic projections, suppress the interference of irrelevant structures, and improve lung nodule detectability. For this purpose, a collaborative detection network is designed to localize lung nodules in 2D dissected projections and 3D physical space. Our experimental results show that our approach can significantly improve the average precision by 20+% in comparison with the common baseline that detects lung nodules from original projections using a popular detection network. Potentially, this approach could help re-design the current X-ray imaging protocols and workflows and improve the diagnostic performance of chest radiographs in lung diseases.