Research And Application On3D Colon Centerline Extraction Algorithm Based On An Approximate Minimum Distance Field

Colon cancer is the third cancer to cause death in the world. Most colon cancers are developed from colon polyps. Detecting and excising the polyps in time will prevent from colon canser effectively. Virtual colonoscopy is a novel tool for detecing colon polyps, it uses the CT or MRI image data of patient’abdomen to reconstruct the3D structure of the whole colon lumen, and detects abnormal morphological changes such as protuberances and polyps by wandering the virtual colon lumen. Virtual colonoscopy is invasive, thus it eases the discomfort of patient during detection to a large extent. And with the introduction of multi-detectors CT recently, CT processing speed and CT detecting-polyp sensitivity are improved greatly. Furthermore, virtual camera can move in any directions to detect the inner wall of colon, thereby increasing the diagnosis effectiveness. However, among the various techniques of virtual colonoscopy, how to determine the wondering path of the3D colon data is one of the difficult problems currently. Presently the most commonly used method is extracting the centerline of colon as the wondering path.There exists various methods to extract centerline at home and abroad, mainly including methods based on topological thinning, methods based on distance transform, methods based on level set. These methods have their own advantages and disadvantages. This paper proposes an algorithm for extracting centerline based on approximate minimum distance field. Each interior voxel in the volume data is encoded with an integer code according to its relative distance from the object border to form an approximate minimum distance field. Then cluster is defined as a set of geometrically connected local maximum voxels with the same distance value. And all the clusters are connected with the shortest paths to form skeletonal points. Then the skeletonal points are distance-coded again to find out the centerline. The proposed algorithm is simple, and the results acquired by the algorithm on an experimental3D colon data demonstrate its efficiency. The proposed algorithm belongs to distance tramsform based method. Before describing the proposed algorithm, we introduce the concepts, theories and algorithms about skeleton extracting based on distance transform in detail. Then we reconstruct3D colon data from2D colon CT images using MATLAB. The reconstructed3D colon is provided as the experimental data. Finally we employ the idea of approximate minimum distance field to extracting2D skeleton of2D binary image, and the results acquired by the algorithm on an experimental data demonstrate its efficiency.