Human Brain Virtual Endoscopy Based On Volume Rendering

Virtual Endoscopy is the simulation of endoscopic interventions using methods of virtual reality, computer graphics and digital image processing. Compared to conventional endoscopy, virtual endoscopy is non-invasive, cost-effective, and free of risks and side effects such as perforation and infection. In additional it is so sensitive that can be used to explore body regions not accessible to conventional endoscopy.Aiming at the task of virtual endoscopy of the human brain, this thesis focuses on the technique of volume rendering based virtual endoscopy of the human brain. The research work can be summarized in the following aspects:Techniques of contrast stretching, median filtering and sharpening are applied to improve the quality of the images got from CT devices.Thresholds of the different objects in the 2D medical images are calculated, using the ICM algorithm for multi-level threshold selection by maximum entropy criterion. Then the cavum in the human brain is segmented slice by slice. Marching Cubes and OpenGL graphic libraries are introduced to realize the 3D visualization.Several automatic path extraction methods in 3D data set, such as 3D thinning, distance transform, slice-based path extraction, optimal reliable path extraction, are researched in this thesis and the algorithm of distance-field based skeletons is implemented. Three modes of a camera control, planned navigation, manual navigation, guided navigation, are implemented.A modified volume rendering algorithm is provided to generate the image. Using the key frames, an animation is generated and adjusted locally to be realistic and smooth.A virtual endoscopy system of the human brain is developed on the basis of the previous work. The correctness and feasibility of the algorithm in this thesis are validated by the system.