Pulmonary fissure segmentation on CT.

Area of Science:

• Medical Imaging
• Radiology
• Computational Anatomy

Background:

• Pulmonary fissures define lung lobes, essential for lobar disease assessment.
• Accurate segmentation of these fissures is clinically significant but challenging on CT scans.

Purpose of the Study:

• To develop and evaluate a novel computational approach for segmenting major pulmonary fissures in both lungs using thin-section computed tomography (CT).
• To improve the accuracy and efficiency of lung lobe segmentation for clinical applications.

Main Methods:

• Proposed a "ridge map" image transformation to enhance fissure visibility on CT.
• Employed a Bayesian network-based curve-growing process incorporating ridge map features and prior fissure shape knowledge.
• Implemented an adaptive regularization framework using an entropy measure to balance feature influences and model dependencies.

Main Results:

• The method achieved high accuracy in segmenting pulmonary fissures, with only 2.4% of regions requiring manual correction.
• The average segmentation distance was 1.01 mm, comparable to inter-observer variability (1.03 mm).
• Segmentation of lung lobes on standard computers was completed in under 5 minutes with linear-time complexity.

Conclusions:

• The proposed method offers an effective and efficient solution for automated pulmonary fissure segmentation on CT scans.
• This technique facilitates accurate lobar-level lung disease assessment and visualization.
• The approach demonstrates robustness and clinical applicability in segmenting lung lobes.