A review on lung boundary detection in chest X-rays

Area of Science:

• Medical Imaging Analysis
• Computer-Aided Diagnosis
• Pulmonary Disease Detection

Background:

• Chest radiography (CXRs) is crucial for diagnosing pulmonary diseases.
• Automated analysis of CXRs requires accurate lung region localization.
• Lung boundary detection is an essential preprocessing step for decision-making algorithms.

Purpose of the Study:

• To provide an overview of recent literature on lung boundary detection in chest X-ray images.
• To review leading lung segmentation algorithms from 2006-2017.
• To identify challenges and publicly available datasets for lung segmentation.

Main Methods:

• Systematic review of lung segmentation algorithms for posterior-anterior and lateral view CXRs.
• Focus on algorithms addressing deformed lungs and pediatric cases.
• Analysis of radiographic measures derived from lung boundaries and their clinical applications.

Main Results:

• Hybrid and deep learning methods show superior performance, comparable to inter-observer variability, but demand extensive training and computational resources.
• Most algorithms are evaluated on standard adult CXRs, neglecting variations in abnormal cases.
• Limited studies exist for pediatric CXRs, which present unique challenges due to developmental differences and image noise.

Conclusions:

• Current lung boundary detection algorithms lack robustness for diverse CXR appearances, including pathological deformities, positioning variations, and background noise.
• Developing algorithms that are resilient to these interferences is a critical ongoing challenge.
• A comprehensive understanding of lung region detection algorithms is vital for advancing automated detection and diagnosis systems for cardiopulmonary pathologies.