Robust Airway Generation Labeling With Airway Segmentation for Reliable Airway Assessment

This study aims to accurately classify the airways in the human respiratory system, which are characterized by diverse pattern forming a complex tree-like structure. Although recent advances in deep learning show promise for automatic airway segmentation, further research is needed to make it practical for determining a patient's airway status for tailored treatment. To enhance diagnostics and treatments, it's crucial to delve into the study of airway generation. This exploration helps in understanding specific structures and pinpointing issues like airway narrowing and lung compliance. Current methods, such as template-matching and machine learning based methods that interpret this as classification problem, have limitations in capturing the complexity of higher-generation airways. To overcome these challenges, our study proposes a novel approach. The Prim algorithm initially establishes a minimum spanning tree from the centerline to create an accurate tree structure that reflects airway connections. A new secondary branch is formed when a main branch with an outdegree of 1 is detected, ensuring precise airway generation labeling. To mitigate false branch generation, the study proposes an approach that increases the cost of trunk lines connected to a centerline with an outdegree of 1. Additionally, a method for pruning trees based on subtree length is proposed to effectively handle segmentation results in deep learning. This method prevents the generation of false branches by removing vertices with shorter subtree airway lengths than their siblings. The approach successfully addresses the common issue of false branch generation, providing reliable labeling of airway generation in higher-generation sections. In conclusion, the technique we propose offers substantial benefits for patient health monitoring, disease prediction, and prevention. It achieves this by providing precise and dependable identification of airway generations within the intricate anatomy of the respiratory system.