Knowledge-based method for segmentation and quantitative analysis of lung function from CT

In this paper we demonstrate that a knowledge-based approach can be used to automatically segment computed tomography images for accurate determination of lung function parameters. Knowledge of the expected size, shape, topology and X-ray attenuation of anatomical structures were stored explicitly as features in a model. These features were used to guide 3-D segmentation of the thorax, trachea, large airways, and left and right lung parenchyma. Lung volumes were calculated from 12 segmented volumetric scans and were strongly correlated with total lung capacity from body plethysmography (r=.95). Dynamic lung attenuation and cross-sectional area changes were calculated from II single-slice flow series during a forced expiratory maneuver. There was no significant difference between these parameters and values calculated from a manually-edited segmentation.