Centerline Extraction Algorithm Of Colon Lumen Based On CT

Colon is one of the important organs with'tubular'structure. Computed Tomography Colonography (CTC) is most commonly imaging method for colorectal diseases screening. Furthermore, image based visualization techniques can assist doctors to view the colon inter surroundings, and extend their vision fields. One of the crucial components of visualization is to extract the centerline of colon lumen. In this thesis, the algorithms of centerline of colon extraction are investigated. The thesis includes three main contents:(1) Volumetric data extraction of entire colon lumen from abdominal CT image series. First, we use double threshords and vertical filter method for 2-D segmentation, and then utilize region growing to extract 3-D colon lumen data form processed binary image series.(2) A fast centerline extraction algorithm based on Maximal Spanning Tree (FMST). The proposed FMST aims to speed up traditional Maximal Spanning Tree (MST) algorithm. FMST cuts off the boundary voxels which do not affect the centricity of the centerline based on search strategies such as Distance From Boundary Primary and Distance From Sourcepoint Secondary (DFBP and DFSS).(3) A testing database of 20 simulated analytic colon models, and a set of quantitative evaluation criterion. We make a comparison among Dijkstra shortest path algorithm (DIJ),MST and FMST.The features of this thesis are:We propose a speed-up version of MST called FMST. Moreover, we generate 20 synthetic geometric models to simulate the colon structure features, i.e. tubular shape, ellipse lumen and high curvature. We also establish a set of quantitative criteria for evaluation.The experimental results on 2 colon CT image series and 10 selected simulated models show that, FMST is much faster than MST (speed up 80%) while preserves the accuracy of centerline (overlap rate 96.98%)...