|
In recent years, medical imaging technology has developed rapidly, healthinformation technology, for example,the CT scanning equipment, the MRI nuclearmagnetic resonance imaging equipment, the ultrasonic equipment, the digital vascularimaging equipment, nuclear medical imaging and so on also have got a great boost.With the hardware technology constant innovation, the software technology alsoneeds to develop itself to work together with the hardware. On the other way, theseimaging equipments above need the help of the supporting software to play its realvalue on the medicine. Currently, the triple reconstruction technique based ontwo-dimensional CT scanning image has grown up. Therefore,3d reconstructionresearch is put on the agenda.Now,3d medical image reconstruction technique is a cross subject and involvesother areas of technology, it applicants with medical knowledge, computer visiontechnology, computer generated imagery, digital image processing techniques, theman-machine interactive technology, etc. But with the development of medicalequipment, there are more and more medical image data, so, it is difficult to renderthese huge data in real time and the image processing module is also need to befurther improved. For the above problems, this article improves the algorithmcombining with GPU hardware acceleration, and obtains the following severalresearch achievements:Firstly, cutting plane method of three-dimensional interpolation based on matrixoperations is proposed. In the multi-planar reconstruction technique, the currentthree-dimensional interpolation methods are very difficult to adapt to the case of data whose amount is increasing. The structure of this method is simple, and it is very easyto implement, the greatest advantage is based on is that it is considered suitable fortransplant to run on GPU; it improved the operating speed greatly.Secondly, segmentation method based on regional statistics and level set isproposed. In medical diagnosis, doctors are very accustomed to want to get the shape,size, and area of ill object that they are interested. For this problem, we proposed theregion-based statistics and level set segmentation methods based on improving theGAC model and C_V model, and this method can be applied to the segmentation ofthree-dimensional planar imaging segmentation, it can realize effective segmentationand extraction.Thirdly, we realized the real-time rendering a great amount of data.Becauseamount of data is too great, it is very difficult to achieve real-time rendering, onlyrelying on CPU, in this paper, acceleration method is based on the OpenGL ShaderGPU hardware, and a large number of data will be processed by the GPU. Becausethe GPU can realize parallel computing, data processing speed on GPU is a hundredtimes faster than the CPU or even thousands of times. GPU’s greatest advantage is itsparallelism. Therefore, the proposed algorithm can be ported to the GPU to realizegeneral purpose computing, and real-time rendering of a large amount of data.Last, we made a friendly human-computer interaction model. According topeople’s daily habits, we proposed a three-dimensional interactive model-ArcBallmodel. The basic idea of ArcBall model is to translate two-dimensional mousemovements to the exercise movements of the control point in three-dimensionalspace, the model realized the three-dimensional reconstruction of human-computerinteraction mechanism, the basic features include drag rotate, pan, zoom. Makehuman-computer interaction becomes arbitrary, handy.
|