Optimizing parameters of an open-source airway segmentation algorithm using different CT images

Area of Science:

• Medical Imaging
• Pulmonary Medicine
• Computer-Aided Diagnosis

Background:

• Computed tomography (CT) is crucial for diagnosing lung diseases.
• Accurate airway segmentation from CT images aids in lung disease assessment.
• Existing airway segmentation methods lack reliability across diverse CT scan parameters.

Purpose of the Study:

• To develop a simple, reliable, semi-automatic airway segmentation algorithm for CT images.
• To optimize the algorithm for consistent performance across various CT acquisition parameters.
• To establish an open-source platform for airway segmentation and evaluation.

Main Methods:

• A region-growing approach is used for segmenting tracheal and bronchial anatomy.
• Independent segmentation of trachea, right, and left bronchi with distinct thresholds.
• Optimization and validation on clinical cases, EXACT'09 data, and a lung phantom with varied CT parameters.

Main Results:

• Successful segmentation across all tested cases with minimal leakage.
• Comparable results to existing methods on clinical data.
• Demonstrated reliability and stability across multiple CT platforms and acquisition parameters, with slice thickness being the most influential factor.

Conclusions:

• The developed system is the first open-source airway segmentation platform.
• The quantitative evaluation method offers a repeatable tool for comparing segmentation platforms.
• The algorithm is stable across CT platforms and acquisition parameters, serving as a foundation for advanced airway segmentation.