Recognition of lung lobes and its application to the bronchial structure analysis

This paper describes a method for recognizing the lung lobes and its application to analysis of the bronchial structure. Analysis of the lung structure is one of important functions in a computer aided diagnosis system for chest CT data. Since the lung is composed of five lobes, analysis of the lung requires recognition of each lobe area. Thin membranes, called interlobar pleura, exist between lobes. Their CT values are higher than those of the lung parenchyma on CT images. Therefore, the proposed method extracts interlobar pleura regions and interpolates the regions by fitting quadratic surfaces. Then, lung regions are divided into lobes using fitted surfaces. From the obtained lung lobe regions and the bronchial tree data extracted beforehand, each bronchial branch is classified into the lobe to which it belongs. The proposed method was applied to fourteen cases of 3D chest CT images. The experimental results showed that lung regions were satisfactorily divided into lobes and that most bronchi were classified into lobes to which they belong