Research On The Path Planning Of Virtual Endoscopy Based On Distance Transform

With the rapid development of medical imaging technology, computer graphics andvisualization technology, Virtual Endoscopy (VE) appeared. VE is widely applied in themedical field and has great significance for medical diagnosis and surgery due to itsnoninvasive examination means. Due to the complexity of the nasopharynx, the research onVirtual Nasopharyngeal Endoscopy is a difficulty and has great application value.The key technologies of Virtual Nasopharyngeal Endoscopy are2D medical imagesegmentation,3D reconstruction, path planning and virtual roaming. In this paper, theresearch focuses on the path planning. Firstly, the one-class Immune Feature WeightedSupport Vector Machine (IFWSVM) is used to segment nasopharynx in MR images aftercomparing its segmentation accuracies with those of the one-class Immune Support VectorMachine (ISVM). Secondly, the fast Marching Cubes (MC) algorithm is used to reconstructthe nasopharynx after comparing its time complexity with that of the traditional MCalgorithm, and the visual effect of the3D model is improved. Thirdly, the DistanceTransformation (DT)-based algorithm is used for the center path and multiple branchesplanning. Two improved path planning algorithms are put forward due to the complexity ofanatomical structure of the nasopharynx and the clinical needs. The experiments indicatethat the DT-based algorithm and the two improved algorithms do well on path planning.Finally, the three times Bezier curve smoothing method is used for3D path smoothing.Three kinds of virtual roaming are realized corresponding to the above three path extractionalgorithms, and a few continuous roaming rendering is presented.The contributions are as follows:1、A selection of branch path algorithm is proposed. The junction points are savedwhen searching for the ending points, so that one of the two or more paths can be chosenwhen meeting a junction point while extracting the path.2、An extraction algorithm for complex path based on interaction is proposed. A centerpath can be extracted by choosing a point through the human-computer interaction. By thisway, the length of the path is arbitrary and there is no need to search the ending points.