Medical Ct Image 3d Segmentation And Visualization

Rapid advances in the field of medical imaging are revolutionizing medicine. Computedtomography (CT), magnetic resonance imaging (MRI), Ultrasonography(US) and other imagingmodalities have been widely used in clinical diagnosis and therapy. With the aids of computers,extracting regions of interest (ROI) from CT, MRI, US and displaying them through 3Dvisualization techniques can help the physician observe the pathological regions more directlyand mo re clearly, which will contribute to improving the accuracy of diagnosis.Image segmentation is one of the most important steps in medical image processing andanalysis. Due to the diversity and complexity of medical images together with noise interference,it is somewhat difficult to achieve an exact segment result. Although many methods have beenproposed to solve the problem during the past 40 years, some of them still have deficiencies inperformance, robustness and automation, which are not suitable for clinical applications.Therefore, designing effective, robust and automatic methods for medical image segmentationhas become the aim that many researchers are pursuing.Medical image visualization techniques play an important role in medical image processingand analysis, and it is one of the most successful applications of visualization in scientificcomputing. Now it has become a useful tool to assist diagnosis and therapy. There are manyvisualization techniques, such as volume rendering, surface rendering and local displaying. Howto make full use of these visualization techniques to help the physician get the most valuableinformation has become a hot issue in visualization research.In this thesis, some work has been done on 3D segmentation and visualization of CT. The maincontributions are described as follows:1. A Framework for 3D CT segmentation is designed and two segmentation methods of 3D CT are proposed.The framework designed in this thesis mainly consists of image pre-processing, 3Dsegmentation and post-processing.(1) In the pre-processing stage, 3D volume is formed by stacking the slices firstly. Then anisotropic diffusion filtering on the 3D volume is discussed. Finally, a tri-linear interpolation algorithm is performed to enhance the inter-slice spatial resolution.(2) In the segmentation stage, a single object 3D adaptive interactive segmentation algorithm (SO3DAISA) based on 3D region growing is proposed for CT segmentation. A growing criterion that combines global and local features is derived and segmentation is performed iteratively to acquire a satisfactory segment result. A criterion that detecting over-segmentation is proposed to solve the over-segmentation problem in 3D region...