Realization Technology For Three Dimension Visualization Based On Medical Slice Images

The research project is one part of "The Diagnostic and Therapeutic System ofComputer Integration", which is supported by "The Tenth Five Years ConstructiveFinance" in college of automation engineering, nanjing university of aeronautics andastronautics. My job is to realize the accurate 3D visualization, fast rendering ofmedical images and their regions of interest. The accuracy and real-time rendering of3D visualization are the powerful guarantee of doctors when they diagnose and treatdisease or make surgical plan, and are the important foundation of implementingcomputer assistant diagnosis and therapy. The main research works and innovativeresults in the thesis focus on the following aspects: 1.A new method of measure and compute the delaying scan time accurately inangiography is proposed. The technique can reduce system noise in imaging processand improve the quality of slice images greatly by using MR fluoroscopy and usingan elliptical centric view order, which can provide a high degree of resolution. Thismethod is progress in technique, and the correlative research is not appeared presently.The scanning slice images with DICOM format are read in personal computer and aretransformed numerical image formats that are convenient for computer manipulation. 2.The wavelet enhancement algorithm and new wavelet multi-scale edgedetection algorithm are introduced. These algorithms can solve the contradictionexisting in traditional algorithms between strong denoise abilities and feebleinformation preservation. The novel algorithm of multi-scale edge fusion is applied tocontrary wavelet transformation, which can implement connection of edge pointwhen we detect edge, and then high-orientation precision edge contour with singlepixel is obtained. Based on edge detection, the seed-fill algorithm, swell algorithmand small-region area elimination algorithm are used to extract the interesting tissuesor organs in single slice. 3.Aim to the problems of powerful computing capacity and slow renderingvelocity, this paper synthetically analyzed all speedup technique of volume rendering,then a re-sampled speedup algorithm of ray casting volume rendering is proposed.Firstly, this speedup algorithm reduced matrix computation by the matrixtransformation characteristics of re-sampling points in two coordinate systems.Secondly, the projection of 3D datasets on image plane is used to reduce the numberof rays. Thirdly, utilizing boundary box technique avoids the sampling in emptyvoxel and greatly improves the efficiency of ray casting. The speedup algorithmovercomes the limitation existing in traditional algorithms that have not a distinctimprovement in rendering speed. This algorithm can produce the required qualityimages and improve the rendering efficiency remarkably. 4.In the field of voxel segmentation, an improved active contour modelalgorithm is developed. In the improved algorithm, a new constitution method ofexterior energy function is proposed. A gray model is applied in edge points of activecontour model, at the same time the grads information of edge points is used tofurther modify exterior energy function. By automatically adjusting the powerparameter, contour line will exactly approach the true object edge. The theoreticalanalysis and experimental results all show that the new algorithm can not only reflectthe characteristic of object edge but also eliminate the influence of noise and falseboundary, as a result the extracted volume will reflect the target body exactly morethan traditional algorithms. 5.A new magnifying algorithm based on wavelet is proposed. In the newalgorithm, the original image is regarded as low frequency of magnifying image, andappropriate high frequency component is constructed to implement contrary wavelettransformation. So as to the blur in magnified images can be eliminated and thedefinition and resolution of magnified images can be improved. Finally, this paper summarizes the works and research fruits and presents thefurther research plan and objective.