Energy Minimization Based Segmentation And 3-D Visualization For Abdominal CT Image

With the rapid development of medical image technology in the medical field, medical image analysis and processing become more and more important and have wide application. Medical image segmentation is used to obtain ROI, organs or tissues from medical images and it is the key prerequisite of follow-up processing. Energy minimization based medical image segmentation method is one of the hotspots in medical image segmentation researching, it deduces the segmentation result by optimizing the energy function constructed by the image information. 3-D visualization of medical images is another important part of medical image processing, it compensates the shortage of transitional medical imaging equipments and could provide realistic 3-D medical images to doctors. The current 3-D visualization of medical image methods mostly focus on surface rendering and volume rendering. Besides, the GPU accelerated 3-D visualization is another research hot spot.This paper focused on the abdominal CT medical image segmentation and 3-D visualization, the main work and innovation are as follows:(1) Introduced the format of medical imaging and related theories, researched the window transform technology of medical CT images, and achieved the analysis and display of standard DICOM format of CT images.(2) Researched image segmentation methods based on implicit and explicit boundary representation, analyzed some common image segmentation principles and implementation methods based on energy minimization, such as Snake model, Balloon model, GVF model and geodesic contour model (GAC), etc. Illustrated the advantages and disadvantages of each method with simulations and experiments.(3) Proposed a two step abdominal CT medical image segmentation method based on improved graph cuts and level set method, this algorithm combined the advantages of graph cuts and level set thus avoided the problems such as graph cuts could not cut into the recessed area of the border and level set could induce leakages, etc. With extending to entire CT image sequence, this algorithm could achieve automatic segmentation of particular organs.(4) Analyzed the intrinsic link of energy functions between graph cuts and level set in depth based on (3), verified the equivalence between the minimization of GAC energy function and minimum cut of graph cuts, on the basis proposed another improved segmentation algorithm which integrated both level set and graph cuts, it used graph cuts to do GAC energy function optimization and simplified the algorithm steps, while it well overcomes the traditional GAC's easy to converge to local minimum problems.(5) Introduced the common 3-D medical image visualization techniques which include surface renderings and volume renderings. Besides, introduced the latest CUDA programming technology, on this basis achieved 3-D texture mapping algorithm and ray-casting algorithm of abdominal CT images. At the same time, used CUDA technology to GPU acceleration of ray-casting and added pseudo-color to improve display effect.