Paper Overview
- Field: Computer Vision (medical imaging)
- Authors: Anna Mrukwa, Marek Socha, Aleksandra Suwalska, et al. (12 authors)
- Published: 2026-08-17
- arXiv: 2608.16855
- Paper: https://arxiv.org/abs/2608.16855
Summary
Background: Lung cancer remains the deadliest cancer worldwide because it is often diagnosed too late. Effective treatment depends on detection at an early screening stage. However, the growing number of patients and the limited number of radiologists lead to prolonged diagnostic waiting times. In very early stage lung cancer, nodule visibility is further reduced by adjacent blood vessels and airway walls, because nodules are often connected to or supplied by these structures. Task-specific analysis of the bronchovascular bundle is therefore important for efficient nodule detection, and its removal can increase the diagnostic potential of lung cancer screening.
Methods: The authors used series from widely utilized low-dose CT (LDCT) datasets, including the Duke Lung Cancer Screening (DLCS) dataset and the Pomeranian lung cancer screening pilot project. The proposed RONALD pipeline operates on CT images and returns binary masks of vessels and bronchi located within the lung parenchyma.
Results: The pipeline segments the bronchovascular bundle in LDCT scans while improving nodule retention rates — on DLCS, from 93.98% and 90.36% to 100%; on the Pomeranian dataset, from 83.16% and 62.36% to 99.92%.
Conclusion: The resulting segmentations can improve pulmonary nodule detection in very early stage lung cancer, and the work demonstrates that unsupervised methods can outperform supervised deep learning in this task when annotated ground truth is limited.