Key Technology Research Of Virtual Colonoscopy Navigation Path And Preliminary Design Of Aided Detection System

With the advancement of computer graphics and medical imaging technology, virtual endoscope (VE) has been an active research field in recent years. Comparison with the traditional optical colonoscopy, virtual endoscope has the advantages of no-pain, no puncture risk, suitable for different patients, and mainly uses in aided detection treatment planning, lesion accurate localization and operation training, etc. The virtual endoscope is mainly applied for the lumen organs, such as ear-nose-throat (ENT), uterus, and colon. In this paper, we focus on colon navigation path approach, which is the key technology of the virtual colonoscopy, and finally design a preliminary virtual colonoscopy system.First, in this paper, we summarize the virtual colonoscopy system structure and technical composition, and then work on three key technologies of virtual colonoscopy, including colon visualization, automatic extraction of centerline and interactive virtual colonoscopy navigation. Finally combining with these three key technologies, a preliminary virtual colonoscopy system is constructed.Colon visualization is the first step of virtual detection. At present, there are two kinds of rendering method usually used in the visualization, which are surface rendering and volume rendering. The Visualization Toolkit (VTK) is an open-source software system for3D computer graphics, image processing and visualization which has integrated both of rendering methods. Therefore, VTK is now commonly used as a research and development tool for virtual reality. VTK also supports a wide variety of visualization algorithms including:scalar, vector, tensor, texture, volumetric methods etc, and it has advance modeling techniques such as:implicit modeling, polygon reduction, meshes smoothing, cutting, contouring, and Delaunay triangulation etc. VTK has so many advantages and provides great convenience for the system development, therefore, in this paper, we introduce the basic technology and characteristics of VTK in second chapter, and particularly introduce the surface rendering algorithm and ray casting volume rendering algorithm. The surface rendering method is finally employed to render the entire or part of the colon for visualization.The centerline extraction of the virtual colonoscopy is a key section in this paper. To guarantee real-time property of the virtual colonoscopy detection system, the centerline extraction should be extracted quickly. In the third chapter, the comparison of a couple of current existing centerline extraction algorithms, such as, manual designated center points, the biggest cost spanning tree method, Dijkstra shortest path method, topology refining method, the level set method, particularly introduced the distance from boundary field (DFB field) is presented. Specially, the biggest cost spanning tree centerline extraction algorithm based on DFB field is analyzed in detail, which can keep the colon topology structure well and easily calculate. However, due to folds, bending characteristics of colon, this centerline extracted by this approach also has too many inflexions, which leads jumping and rotating during navigation vision. At the same time, because of colon’s long length, the centerline with too many center points will lead to a long time calculation and finally increase the center path generation time. To get a smooth navigation path and shorten the centerline extraction time, based on the DFB field algorithm, we propose a center points decrease algorithm which reduces the number of center points and improved the path construction speed. Then we use the cubic-spline interpolation algorithm to smooth the centerline, eventually reduce the rotating and jumping during navigation vision. The smooth and fast colon centerline extraction is achieved finally.After extracting the centerline, the next step is to choose the fitting navigation observation model for colon inspection and pathological detection.In the fourth chapter, we firstly summarize some usually used navigation observation methods:fly-through mode navigation inside the colon, eversion and rotation navigation and virtual flattening display. The traditional fly-through mode navigation display technology simulates the traditional optical physics endoscope in the observation methods. It is a kind of the earliest display technology, which has the shortcomings of narrow visual space and big invisible region. The method of eversion and rotation replaces the outside colon wall with the inside colon wall, so make observe the colon internal structure easier. However, this method has a drawback that in colon bending area, there is easy to appear more or less turn over. Virtual flattening navigation method is to virtually divide and flat the entire colon into a2D image, although this method increases the view area, but the method destroyed the whole cavity viscera anatomical configuration, makes the observer hard to imagine the structure of the original viscera and is not fit to actual application. In this paper, we develop a colon section and split approach. The colon is divided into many small sections along with the centerline, and each section is split into two halves next. Finally, a virtual camera is posed above the split sections to observe. This method could perfectly observe the internal structures between the colon folds, and decreases the invisible area obviously. Meanwhile, a second section approach is used to overcome the mis-division of colon, which is based on curvature of centerline calculation.In the fifth chapter, a preliminary virtual colonoscopy system which integrates colon render, centerline extraction and navigation observation is introduced. The system is developed by C++graphical interface framework Qt which supports cross-platform, adopts the idea of object-oriented programming, performs the pipeline flow method through a procedure of data input, data transformation, DFB field construction, centerline extract and display, the first and the second colon division, overlooking navigation and detection. Meanwhile, the system also develops a manual navigation mode, with the interface control buttons and mouse interactive operation on the render interface. A system with basic colon aided detection function is finally achieved.At last, the development of virtual colonoscopy is discussed in chapter six, and the future work is introduced also.