An extraction algorithm of pulmonary fissures from low-dose multi-slice CT image

Recently, the development of multi-slice CT scanner makes it possible to precisely measure whole lung area in short period time. The improvement of the spatial resolution along z-axis requires the development of three-dimensional (3-D) image processing techniques. This paper presents an algorithm to extract pulmonary fissure part, which is one of important organs to identify lung area, from multi-slice CT images. Since the pulmonary fissure has extremely thin sheet-like structure, its extraction is one of challenging problems. This algorithm mainly consists of two procedures, identification of the region with high likelihood of fissures, and enhancement of the thin sheet-like structure. The identification procedure utilizes the 3-D Euclidean distance from pulmonary vessels which are classified into each lung lobe area. For the enhancement procedure, a filter that responds to the sheet-like structure was designed. Once these procedures performed, a region growing technique based on the normal vector of the enhanced sheet like structure is applied to extract the fissure region. The usefulness of the proposed algorithm is demonstrated by using twenty clinical data sets. (author)