Visualization of large medical volume data

Abstract

Visualization of large medical datasets is important in the field of computer aided surgery. It is a challenging problem that requires a solution integrating image processing and computer graphics. This thesis investigated various methods of 4-D medical image compression including clustering-based compression, and a new dynamic data compression method using motion compensation and hierarchical vector quantization. This method can achieve a higher fidelity and faster decompression time compared to existing methods with comparable compression ratios. With real-time decompression, we are able to achieve interactive rendering of very large 4-D medical images. The thesis also addresses the problem of interactive rendering of medical images with a new clustering-based automated transfer function generation algorithm that makes use of the image features in the dataset. The image-based visualization solution which we investigated in this thesis is demonstrated in an application on computer graphical simulation of vascular drug flow.