LTSP: long-term slice propagation for accurate airway segmentation

Area of Science:

• Medical Imaging
• Pulmonary Medicine
• Computer-Aided Diagnosis

Background:

• Bronchoscopic interventions for pulmonary diseases rely on accurate airway maps.
• Chest computed tomography (CT) scans are used for airway segmentation.
• Maintaining airway topology completeness during segmentation is challenging.

Purpose of the Study:

• To develop an accurate airway segmentation method for pathological CT scans.
• To improve the localization and guidance for bronchoscopic interventions.
• To address the challenge of preserving airway topology completeness and segmentation accuracy.

Main Methods:

• Proposed a long-term slice propagation (LTSP) method for airway segmentation.
• Designed a two-stage end-to-end segmentation framework.
• Utilized LTSP in the decoding process to exploit slice continuity and enhance weak airway features.

Main Results:

• The LTSP method demonstrated considerable improvements on 70 clinical CT scans.
• Most airway breakages were eliminated, and more tiny bronchi were detected compared to state-of-the-art methods.
• Ablation studies confirmed the effectiveness of the method's components and framework design.

Conclusions:

• Slice continuity information is crucial for accurate airway segmentation.
• Propagating long-term slice features preserves airway topology connectivity.
• The proposed method maintains overall segmentation accuracy while enhancing topological completeness.