A Fully Automatic Segmentation Method for Virtual Bronchoscopy

This paper presents a fully automatic segmentation method for the virtual bronchoscopy, which requires the segmentation of a trachea and left/right bronchi. A trachea region is automatically detected, based on medical knowledge and DICOM header information, and its center point is selected as a seed point for the slice-based 3D Seeded Region Growing(SRG). The slice-based 3D SRG grows a region by collecting its homogeneous neighbors in a single slice, and makes seed points a t adjacent slices. The seed points grow a t their own slices, recursively. While growing, a region size constraint between adjacent slices is preserved to prevent from leakage problem of conventional SRG methods. The slice-based 3D SRG can segment a complex tubular shaped organ whose center axis can be represented by a high order curve. Segmentation results are presented using CT slice image data sets.