Session PB7.8

Visualization of Segmented Cardiac Anatomy with Accelerated Rendering Method

F Yang, WM Zuo*, KQ Wang, HG Zhang

School of Computer Science and Technology
Harbin City, China

Volume rendering of 3D anatomical and medical data would be valuable in medical diagnosis and surgical planning. Using the cross-sectional data from the Visible Human Project, we investigate the visualization of the segmented female heart data from three aspects, transfer function design, interactive visualization, and accelerated rendering. In transfer function design, we assign an appreciate opacity and color value for each kind of cardiac tissue to provide a satisfactory visualization quality on the shape and boundary of tissues. In interactive visualization, we adopt the ray casting algorithm of volume rendering, where the result image is rendered and each significant tissue of the heart is clearly displayed. This rendered virtual model is important for characterising variant tissues and help us to understand the link between the structure and the electrophysiological function of the heart. We also provide the fundamental rotation and zooming operations in the visualization platform. We further implement an interface to display three regular views of the volume data, known as coronal view, sagittal view, and transverse view. With these three regular projection views used in medical diagnose, different tissues are corresponded to their respective regions in slices of volume dataset. Furthermore, a clipping plane tool is provided to crop the heart at an arbitrary point of view. Irregular cropped sections of an anatomy model are constructed. Using this method, the interior structure of the heart under this condition can be displayed. In accelerated rendering, we present an accelerated rendering method to speed up the original ray-casting rendering method. In pre-processing of the segmented dataset, boundary voxels and non-boundary voxels are identified and the non-boundary voxels are labeled with a value out of the value range of the segmented dataset. When casting a ray through the volume, those interior voxels of each tissue are treated together and the color and opacity value of boundary voxels of tissues are mapped from densities as in a normal ray-casting algorithm. Since there are a dominant proportion of non-boundary voxels in the segmented dataset, this method will efficiently improve the expensive computing problem of ray casting.

(Abstract Control Number: 212)