Iso-Surface Volume Rendering : speed and accuracy for medical applications
Marco K. Bosma · University of Twente Research Information · 2000
This thesis describes the research on the accuracy and speed of different methods for the visualization of three-dimensional (3D)sets of (measured) data. In medical environments, these 3D datasets are generated by for instance CT and MRI scanners. The medical application makes special demands on the visualization methods. Medical application of 3D visualization methods asks for high accuracy. For practical application, the speed of the visualization should also be high, preferably real-time (25 images per second or higher). Because most 3D visualization methods are very computational intensive, real-time high quality visualization is not (yet) possible in software. As a consequence, many different visualization methods have been developed that either sacrifice the speed for quality or sacrifice the quality to obtain high speed. To meet both the demand for high speed as well as the demand for high quality, a visualization method has been developed that reduces the number of computations drastically without affecting the image quality. This visualization method, called Iso-Surface Volume Rendering, is able to find and visualize surfaces in a threedimensional dataset with high speed (approximately 5 images per second). To be able to compare the accuracy of different isualization methods, the signal theoretic background of the isualization methods and the data acquisition process for CT and MRI scanners is investigated. Using this information, the accuracy of the approximation of the location of a measured step edge is investigated. This accuracy only depends on the parameters used for acquisition, and is hence independent of the visualization method used. It is possible to approximate the location of the step edge using a surface at which the measured data has constant value. This value is called the iso-value and the corresponding surface is called the iso-surface. Next, the accuracy at which the different visualization methods are able to determine and visualize the location (and shape) of this isosurface is investigated. Some methods are not able to visualize a surface, while the accuracy of other methods is limited because the data values are only determined at some fixed locations. The accuracy of these methods can be improved by decreasing the distance between these points. This will however also increase the calculation time dramatically. The Iso-Surface Volume Rendering method uses iteration to find the location of the iso-surface with a very high accuracy, without significantly affecting the calculation time. Finally, a description of the different optimizations used to increase the speed and some examples of application of the Iso-Surface Volume Rendering on medical data are given.