Publications

Efficient Volume Visualization of Large Medical Datasets

S. Bruckner

Abstract

In volume visualization, huge amounts of data have to be processed. While modern hardware is quite capable of this task in terms of processing power, the gap between CPU performance and memory bandwidth further increases with every new generation of CPUs. It is therefore essential to efficiently use the limited memory bandwidth. In this paper, we present novel approaches to optimize CPU-based volume raycasting of large datasets on commodity hardware. A new addressing scheme is introduced, which permits the use of a bricked volume layout with minimal overhead. We further present an extended parallelization strategy for Simultaneous Multithreading. Finally, we introduce memory efficient acceleration data structures which enable us to render large medical datasets, such as the Visible Male (587x341x1878), at up to 2.5 frames/second on a commodity notebook.

S. Bruckner, "Efficient Volume Visualization of Large Medical Datasets," in Proceedings of CESCG 2004, 2004.
[BibTeX]

In volume visualization, huge amounts of data have to be processed. While modern hardware is quite capable of this task in terms of processing power, the gap between CPU performance and memory bandwidth further increases with every new generation of CPUs. It is therefore essential to efficiently use the limited memory bandwidth. In this paper, we present novel approaches to optimize CPU-based volume raycasting of large datasets on commodity hardware. A new addressing scheme is introduced, which permits the use of a bricked volume layout with minimal overhead. We further present an extended parallelization strategy for Simultaneous Multithreading. Finally, we introduce memory efficient acceleration data structures which enable us to render large medical datasets, such as the Visible Male (587x341x1878), at up to 2.5 frames/second on a commodity notebook.
@INPROCEEDINGS {Bruckner-2004-EVV,
author = "Stefan Bruckner",
title = "Efficient Volume Visualization of Large Medical Datasets",
booktitle = "Proceedings of CESCG 2004",
year = "2004",
month = "apr",
abstract = "In volume visualization, huge amounts of data have to be processed.  While modern hardware is quite capable of this task in terms of processing  power, the gap between CPU performance and memory bandwidth further  increases with every new generation of CPUs. It is therefore essential  to efficiently use the limited memory bandwidth. In this paper, we  present novel approaches to optimize CPU-based volume raycasting  of large datasets on commodity hardware. A new addressing scheme  is introduced, which permits the use of a bricked volume layout with  minimal overhead. We further present an extended parallelization  strategy for Simultaneous Multithreading. Finally, we introduce memory  efficient acceleration data structures which enable us to render  large medical datasets, such as the Visible Male (587x341x1878),  at up to 2.5 frames/second on a commodity notebook.",
pdf = "pdfs/Bruckner-2004-EVV.pdf",
images = "images/Bruckner-2004-EVV.jpg",
thumbnails = "images/Bruckner-2004-EVV.png",
note = "CESCG 2004 Best Paper Award and Best Presentation Award",
affiliation = "tuwien",
url = "//www.cescg.org/CESCG-2004/web/Bruckner-Stefan/html/"
}
projectidprojectid

Media

Downloads

Full paper [PDF]