Stereoscopic medical imaging
Abstract
This thesis investigates the use of stereo techniques in medical imaging. Two important aspects, namely, stereo visualization and stereo reconstruction of x-ray images, are covered. It is a well known fact that humans perceive the world in three dimensions because of the different viewpoints of the left and right eyes. However, an important question is whether stereo visualization of x-ray images is possible, and if so, whether it would be effective in aiding diagnosis. To answer these questions, we have conducted a systematic receiver operating characteristic (ROC) study. Results of the study demonstrated the effectiveness of stereo viewing in aiding the detection of shape abnormalities in radiographic images. To extract quantitative information, 3D reconstruction is necessary. A hierarchical feature-based stereo reconstruction algorithm for angiography applications is described. Our technique uses a natural decomposition of vascular structures into trees, branches and curves, to solve the stereo correspondence problem. This results in more efficient and globally consistent matching. Besides allowing the extraction of quantitative measurements, the reconstructed model can be viewed from arbitrary direction(s); lighting and shading can also be added for enhanced visualization. To obtain x-ray test images for our ROC study, and for reconstruction testing and visualization of reconstruction results, we have implemented a geometric modeling system using constructive solid geometry and ray tracing techniques. Our system handles a broader class of generalized cylinders than any other reported ray tracer. Using our system, both simulated x-ray and visible light images can be generated. With high quality stereo visualization and improved methods of stereo reconstruction, stereoradiographic approaches may quickly become an integral part of clinical imaging. It is a cost-effective technology which can provide physicians with three-dimensional visualization and quantitative measurements for diagnosis and follow-up care of patients.
Degree
Ph.D.
Advisors
Chelberg, Purdue University.
Subject Area
Biomedical research|Electrical engineering|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.