| Since the seventies, the medical imaging techniques such as the X-ray computer tomography, magnetic resonance imaging, and ultrasound imaging have gotten a rapid development, these imaging technologies are widely applied to the medical device industry, have developed a number of powerful medical imaging equipment such as CT, MRI, PET and so on. People can quickly and easily get the two-dimensional digital tomographic image sequences of the body through medical imaging equipment, then analysis to identify the lesion area, the timely development of the treatment plan. Limited to technical issues, early diagnostic imaging was conducted based on the two-dimensional digital tomographic images, this is visually difficult to clearly determine the internal organs’s three-dimensional structure or reflect spatial relationships. In fact, these two-dimensional digital tomographic image sequence already contains three-dimensional information of the body’s internal organs, how to transform these three-dimensional information into the graphic images which comply with the people’s visual perception, has been a three-dimensional medical image visualization technology focus. With the rapid development of computer graphics, derived from the scientific visualization research, research in this area to make a three-dimensional medical image visualization become a reality. Three-dimensional medical image visualization refers to the medical image data are processed by a computer, to convert it into an image with a three-dimensional effect to show the three-dimensional shape of human tissue, and a human-computer interaction technology. Three-dimensional medical image visualization techniques can usually be divided into surface rendering and volume rendering methods. Surface rendering is the reconstruction of the surface, the basic idea is to extract surface information of interest from slice data sets which contain the three-dimensional data, using a series of polygonal surface of the sheet to fit the isosurface, and then carried out by conventional graphics techniques to obtain a three-dimensional image. Surface rendering can effectively draw a three-dimensional body surface, but lack of expression of the internal details. Thus, in the late1980s that people put forward the concept of volume rendering. Volume rendering process eliminates the need to construct an intermediate geometric primitives, for all volume data processing directly to obtain images with three-dimensional effect. The advantage is you can render directly without segmentation in favor of preserving the details of the three-dimensional medical image information, to enhance the overall effect of rendering, but the drawback is the need for all voxels processing, increased computational overhead, limiting the image rendering speed. With the development of computer hardware, especially graphics performance leap in growth, the dedicated intensive calculation graphics processor-programmable GPU emerges. GPU can handle massive amounts of data with ease due to its full capacity of high-performance parallel computing. For this reason, the volume rendering of three-dimensional medical images has become increasingly popular, accelerated volume rendering algorithm is proposed from different angles by many scientists, which greatly improves the speed of volume rendering has great potential for development. In this paper, according to a progressive manner, first chapter introduces the research background and significance, noting that the main problems faced by the volume rendering has two, one is how to improve the rendering quality, this choice is inseparable with the algorithm itself, different algorithms because the framework is not the same principle, the order of data processing also vary widely, so the results will be different. Further, transfer function design is also a key factor that affecting the rendering quality. The transfer function determines the result of volume rendering, the values of the sampling points of the volume data will be converted to visual attributes-color and opacity which are then converted to an image by fusing, the user can observe different characteristics of the volume data by adjusting the transfer function, for more useful information. Currently, most of the transfer function requires a lot of manual intervention in order to achieve satisfactory results, how to make the transfer function design more intuitive to use is a hot research field of volume rendering. Another problem faced by volume rendering is how to improve the rendering speed in order to achieve real-time interaction, which is the focus of this research work. From a software perspective, the accelerated algorithm includes early ray termination, empty space skipping and so on, the essence of these technologies is to ignore voxels which do not contribute to the final image, reducing computing. From a hardware perspective, we can design dedicated graphics card according to the algorithm characteristics, optimize the data storage structure, improve the efficiency of data access. Among them, the hardware acceleration support is the most effective accelerate strategy. The second chapter describes the basics of volume rendering. Volume rendering technology development so far, algorithms vary, but they all have a basic set of drawing processes, namely volume data acquisition, preprocessing, resampling, classification, fusion, display, and some may also contain shading calculation process. We put this part of the contents in the second chapter, is to allow the reader has a general grasp of the volume rendering process, knowing that what volume rendering process doing in the end, what processing operations are performed on the data, the data between each stage how to pass and outputs the final results. After the reader has basic knowledge of the second chapter, we launched our research work in the third chapter. On the basis of in-depth analysis of the principles of ray-casting algorithm, we found that there are certain correlation between the light and take advantage of this correlation on ray-casting algorithm for secondary transformation, can greatly improve the rendering speed. The main work of this paper is based on a programmable GPU ray-casting algorithm, the classical ray casting algorithm is improved, the use of streaming GPU’s parallel computing architecture, combined with OpenGL offscreen rendering technology and image interpolation algorithm for three-dimensional reconstruction, achieve the effect of real-time display on the ordinary PC. The main concrete work done as follows:1. Because of high rendering quality, less aliasing factor algorithm itself introduced, ray-casting algorithm becomes a hot area of visualization research. But the amount of data processed is particularly large, high computational complexity and computational time, resulting in the algorithm are not widely used. Based on the rays parallel sampling characteristics, we use programmable graphics processor powerful parallel processing performance to accelerate the algorithm, while in the premise of ensuring the quality of the rendering result, achieve a real-time rendering.2. The calculation of the ray-casting algorithm is large due to the number of sampling light which directly increases the complexity of the sampling point interpolation, classification, fusion. The reconstructed image display resolution and sampling rays are closely related that it directly affects the rendering speed, by reducing the image display resolution can significantly improve rendering speed. Therefore, we use OpenGL off-screen rendering technology to control the number of incident light, that is, set the rendering resolution lower than the normal display resolution4-16times, then carry out normal sampling ray casting algorithm and get intermediate results.3. In order to get the normal display resolution images, we will get lower rendering resolution image re-used as input for interpolation reconstruction. In order not to reduce the quality of the reconstructed image in the premise and improve reconstruction algorithm speed, we use the Catmull-Rom interpolation algorithm as high-resolution reconstruction algorithms, and run on the GPU fragment processor. The experimental results show that this method can meet the prerequisite reconstructed image quality, effectively improve the speed of reconstruction.In addition, the paper also singly introduces the MPPS service of DICOM standard in the fourth chapter. With the constant improvement of digital hospital construction and medical information standardization, the status of the equipment during the examination, inspection methods, image information generated during the scan, the radiation dose and other information materials required for billing have become concerned about the content of clinical. However, most current RIS and PACS systems have no way to get this part of the information from the equipment. The MPPS service of DICOM standard is designed to provide external system with access to the device status information during the inspection. Therefore, the authors deeply study the MPPS service of the DICOM standard, analyze the information content, services timing status and detailed service process, then based on the development kits DCMTK package implement MPPS services. Simulation experiments show MPPS services provide timely and accurate inspection information. |
PDF Full Text Request