Flying Through the Heart
Flying Through the Heart
Scientists at Oxford University are building computer models of the heart in order to further basic science understanding of the heartbeat, and to help make the diagnosis and treatment of cardiac disease more efficient. In the future, the models will be used to simulate alternative surgical scenarios for the patient, such as the re-opening of blocked vessels, or the fitting of stimulation leads, so that surgeons will be able to make more informed decisions about their operating procedure. Experimental data forms part of the input to the scientists' models, and visualization of this data is required for better understanding of relevant structures. One of NAG's Senior Technical Consultants, Dr Jeremy Walton, has set up the visualization at Oxford as part of his consultancy work at the Oxford e-Research Centre (OeRC).
The experimental data underlying this visualization comes from a BBSRC-funded collaboration between the teams of Dr Peter Kohl (Department of Physiology, Anatomy and Genetics), Dr Jurgen Schneider (Department of Cardiovascular Medicine) and Professor David Gavaghan (Computing Laboratory). It comprises a stack of images generated from a sequence of high-resolution magnetic resonance imaging (MRI) scans of a heart, which are then segmented to discriminate between cardiac structures. The visualization software, developed by Dr Christopher Goodyer of the School of Computing at the University of Leeds (as part of the EPSRC-funded Integrative Biology project) calculates isosurfaces through the segmented MRI data, generating 3D surfaces of tissue structure.
The ever increasing resolution of experimental devices such as the MRI scanner is leading to larger datasets (for example, the heart data has a size of 1.5 GB) which cannot be viewed on conventional displays without loss of detail. Accordingly, the tissue surfaces were originally shown on the high-resolution display wall at Leeds (a visualization that won Christopher and his colleague Professor Ken Brodlie first prize in the vizNET Showcase 2008). With the help of the Leeds team, Jeremy ported Christopher's application from their display to the video wall at OeRC, which facilitates its use by the Oxford scientists, and promotes its wider dissemination. (Jeremy has written up a detailed technical account of the porting work here [pdf]).
Figure 1: Using the OeRC video wall for large-scale display of 3D heart data. This view shows the exterior of the heart.
Large-scale displays such as the OeRC video wall are built by tiling together several conventional LCD screens that are powered by a compute cluster. For example, every node in the OeRC cluster is responsible for two screens, each of which has a size of 24 inches and a resolution of 1920 by 1200 pixels. The sixteen screens are arranged in a four by four array, giving a total resolution of 30 million pixels. Programs which use the wall have to handle the distribution of the scene to be rendered across the compute nodes so that each part of the display is coordinated with its neighbours. They also have to respond to commands from user-controlled input devices which allow the user to move the camera viewpoint and interact with the scene. A fixed camera position generates a static view, whilst programming a sequence of camera positions results in a flight through the 3D scene (see, for example, the video featured here). The high resolution of the display means that it can be used to visualize very large datasets such as the 3D heart; typical scenes displayed here contain tens of millions of triangles. Moreover, the physical extent of the display (which, if the user stands close enough, can fill their field of view) leads to a compelling sense of the viewer being "inside" the scene, and also means that fine details can be explored without losing contextual information about location within the dataset. Other aids to understanding and orientation such as text labels and navigational reference points can be incorporated into the visualization.
Figure 2: As Figure 1, but now showing the interior structure of the heart.
The 3D heart visualization has recently been featured on the BBC website, in a story which provides an illustration of the use of video walls for cardiac data display. Future work will concentrate on offering the user further options for navigation through the scene; the application of the video wall to other high resolution datasets will also be explored.