Dissertation > Excellent graduate degree dissertation topics show

Real-Time Interactive Visualization in Medical

Author: WangWenHua
Tutor: FengQianJin
School: Southern Medical University,
Course: Biomedical Engineering
Keywords: Volume Rendering Fusion MPR MIP
CLC: TP391.41
Type: Master's thesis
Year: 2013
Downloads: 25
Quote: 0
Read: Download Dissertation

Abstract

Visualization is a technology using computer graphics and image processing, converting data into graphical or image displaying on the screen. The applications of visualization related to scientific computing, virtual reality, mechanical engineering, auxiliary medical diagnosis and so on. Visualization in medical convert medical data, such as CT, MRI, PET, into visual images through a variety of real-rendering as well as non-authenticity to help doctors analysis complex pathological conditions.I’m intending to describe my job according to the progress of the study and research.Let’s start from the basic volume rendering. Volume rendering is a technology that generates a two-dimensional image from volume data. The common algorithm of volume rendering include ray casting method, shear transformation method, three dimension texture method and footprint method. The most famous algorithm is ray casting in recent years which is based on the absorption and emission model proposed in1989. The main steps of ray casting include classification of volume data, accumulate by the color and opacity and illumination model for enhancing the sense of reality.The largest advantage of volume rendering is the ability of exploring the internal structure of objects and describing very stereotyped objects. The drawback is too large data to cost much compute time. Therefore, the current research focus on volume rendering algorithm based on programmable graphics card and the parallelization rendering algorithm.These are all reality-rendering which try to convert the volume data to real human body model, freeing doctors from the ocean of data, giving doctors the most intuitive visual recognition. There have some problems that some focus areas will be covered by the organization above when doctors observe areas of disease, tiny blood vessels and so on. The reality-rendering may not help. Non-reality-rendering emerged. It means the use of computer-generated a photo which does not have realistic graphics rendering but artistic creation.Non-reality-rendering method used for solving the occlusion problem include cutting, optical model, sculpt, context preserve volume rendering and so on.Visualization helps doctors to obtain complex internal structure of medical data. But different types of medical data often contain different information or even complementary information. For example, PET can provide detailed functional features while CT has structure features which can be used for accurate anatomic localization. What about combine these two things? It comes to find a reasonable fusion method to achieve in the case of non-interfering while expressing the internal information of different data.We use two transfer functions for classification of PET and CT. Then we compute statistical information of PET and CT as the fusion weights. The result contains a variety of information which is accurate, sensitive and precise positioning. Doctors use it to explore patient’s body, find lesions and diagnose.Compared to the complexity of volume rendering, maximum intensity projection (MIP) is a simple method for generating images similar to X-ray. We compare each voxel density along the ray from eye position to the projection plane, and get the largest values for the terminal image.Mean intensity projection (MeanIP) and Minimum intensity projection (MinIP) are similar to maximum intensity projection. MinIP get the smallest values along a plane direction and is mainly used for airway display, or the display of intrahepatic expansion of the bile duct in the liver enhanced as well.Volume rendering generate images for direct visual effects which tend to be consistent with human visual. Doctors use these images for a qualitative analysis. But when a doctor needs an accurate diagnosis, or need to do quantitative analysis such as statistical average gray value, volume rendering doesn’t work. Multi-planar reconstruction (MPR) meets these requirements. MPR maintains the volume density faithfully. A MPR image needs axial scan and overlay, then re-organization designated by recombinant line such as coronal, sagittal, or oblique images at any angle reorganization. The disadvantage of MPR is that it cannot express spatial structure as volume rendering does.A more complex condition is MPR with thickness which combines MPR with MIP. This type of MPR is not just one slice but many slices which is quantified with slice number of slice thickness in millimeters. This approach is often used in the display of vessel. Doctors locate MPR to the vessel and adjust the MPR thickness to show vessel exactly. The thickness in clinical use is between100millimeters and400millimeters.I introduced from the basic volume rendering, to cutting, context preserve volume rendering, sculpt, fusion display and MPR above all based on the clinical. We use programmable graphics cards (Graphic Processing Unit, GPU) to get an interactive frame rate. The advantage of GPU is that it can provide hundred times in CPU performance. But when we come to large volume data, GPU can do nothing because of memory overflow. Effective solutions include re-allocating GPU memory reasonably, or we seek back to CPU acceleration algorithm.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile