Research In 3D Visualization Of Medical Image Based On GPU

Direct Volume Visualization is an efficient technique to explore complex structure within volumetric data. Its main advantage, compared to standard 3D surface rendering, is the ability to perform semitransparent rendering in order to provide more information about spatial relationships of different structures. The huge amount of data sets produced by the mordern imaging equipment present great challenge to current rendering architectures and techniques. The rendering methods in visualization and computer graphics are focusing either on image quality in order to produce high quality images with non-interactive, or sacrifice quality in order to attain interactive or even real-time performance. However, with the evolution of graphics hardware, the researchers have developed many new algorithms which concentrate on both high image quality and interactive performance. The algorithms take the specific structure of graphics hardware architectures into account and have great meanings for real-time 3D reconstruction and visualization technology. In this paper, exploration on visualization technology of medical images have been made meaningfully based on studying GPU programming.1. Fast volume rendering of medical images on GPUThe powerful parallel computational capability and flexible programmability of Graphic Processing Unit (GPU) are made fully use. We implement the traditional ray-casting-based volume rendering algorithm on the programmable graphics processing unit. By the algorithm, we put the time-cost trilinear interpolation and resampling steps in GPU to shorten the time. First, the data sets and the transfer functions are transferred into the video as textures. Then the entry-point and the leave-point of the ray and the compose of the ray are calculated by the vertex shaderand the pixel shader. Finally, different rendering effects are achieved by modifying the transfer function. The technique of rendering to the render target is used in our algorithm. We use textures to keep the intermediate results so that dynamic branching can be avoided. The experiment shows that our algorithm is significantly fast and can meet the requirements of real-time interactive rendering on the premiseof ensuring imaging quality.2. Fast volume clipping of medical images on GPUIn order to provide doctors with comprehensive, visual and accurate diagnostic information, fast volume clipping of medical images on GPU is proposed in the thesis. The rapid volume clipping is achieved by combining the clipping algorithm with the fast volume rendering algorithm based on GPU. The technique is based on fast volume rendering of medical images on GPU. It works as follows: The space information of the section plane is sent to the shader. Then compare the position of the volume data sets with the section plane to decide whether the data is clipped or not. Our algorithm is different from the algorithm that clipping based on depth stencil. We can see the result that reconstructed from the data sets that remained from any position as long as the section plane is defined. According to the shape the users defined, our algorithm can decide which parts of the volume have to be clipped. Because of using the general graphics hardware acceleration, the proposed method achieves interactive display rate and can be used in surgery simulation and so on.3. Fast volume rendering of medical images driven by importance factorDoctors can see the region of interest even in the middle of the volumetric databy volume clipping. However, it is hard to define the clipping plane. If the clipping plane cuts the object of interest then a part of this object is not visible in the resulting visualization. If the clipping plane is placed right in front of the object of interest then some parts of the object will be occluded by regions of the volumetric data. Fast volume rendering of medical images driven by importance factor is proposed in this thesis to eliminate these drawbacks. Besides applying the hardware acceleration, our algorithm inducts an additional value named importance factor in the transfer function space. We assign high value to the parts that are important and interested in and low value for the parts in the contrary. We are able to highlight the important structure and interested organs and suppress the less important structure and uninterested organs by modifying the factors. An image-based lighting model uses sphere maps to represent non-photorealistic rendering styles. By mimicing the style of traditional illustrations, we can make a perfect depiction of important structures and details. This method can make up the insufficiency of the traditional clipping algorithm. It can show the internal structure clearly at an interactive speed and can provide the doctors as much information as possible.