Methodologies for the 3-dimensional display of volumetric data and implementation on a parallel processing systems

• Main Author: Lomax, A.
• Published: University of Aberdeen 1992
• Subjects: 502.85
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.592961

Many imaging modalities available to the clinician are capable of producing volumetric data sets of example, X-Ray CT, MRI, SPET and PET techniques. In this work, methods for the surface rendered display of such data sets have been developed, based on the oblique casting of rays to determine the visible portion of an object. Local surface contrast is enhanced by the application of a diffuse lighting model, in which surface normals to an object are defined either by the use of a simple gradient shading algorithm, or by a more sophisticated technique based on the definition of variable area triangles. View generation times have been reduced by the use of both software and hardware approaches. The software approach exploits the coherence between consecutive views when an object is rotated about a given axis, whilst the hardware approach uses parallel processing methodologies in the form of a transport network. Execution times for the visible voxel detection algorithm have been found to range from 160 - 600 milli seconds when implemented on 13 transputers. When image coherence is utilised this is reduced further to between 130 - 250 milli seconds per view. For the gradient shading algorithm, execution times on the same transputer array ranged from 80 - 100 milli seconds. A particular application of such surface rendering techniques is described for the display of gated cardiac blood pool images, with colour and dynamic display options being used to convey temporal or functional information about the dynamics of cardiac motion.