Knowledge-based segmentation of thoracic computed tomography images for assessment of split lung function

Area of Science:

• Medical Imaging
• Pulmonary Medicine
• Computational Anatomy

Background:

• Assessing differential lung function is crucial for patients undergoing lung resection, such as lung transplantation.
• Current methods may not provide sufficient detail for precise pre-surgical planning.

Purpose of the Study:

• To develop and validate an automated, knowledge-based segmentation algorithm for deriving functional information from dynamic computed tomography (CT) image data.
• To automatically calculate median lung attenuation and area measurements for individual lungs during forced expiration.

Main Methods:

• A knowledge-based segmentation system utilizing an anatomical model and inference engine was employed.
• The system segmented thoracic CT images, identifying chest wall, mediastinum, trachea, airways, and lung parenchyma.
• Automated segmentation accuracy was validated against manual editing by an expert observer.

Main Results:

• The knowledge-based segmentation achieved >98.55% accuracy in classifying lung vs. non-lung pixels.
• Manual editing was required for 13.94% of images.
• No significant differences were found in median lung attenuation or area values between automated and edited segmentations (p > 0.70).

Conclusions:

• Automated, knowledge-based segmentation can accurately derive indirect quantitative measures of single lung function.
• This method offers functional insights not obtainable through conventional pulmonary function tests.
• The algorithm supports pre-operative assessment for lung resection procedures.