Ray based exploration of volumetric data

In this work we present techniques for the exploration of three dimensional (3D) volumetric data that provide a key to the understanding of data. Today, physicians and scientists have access to advanced scanning modalities such as computed tomography (CT) and magnetic resonance imaging (MRI) that produce complex high resolution volumetric datasets. In order to make a timely diagnosis, physicians need to quickly explore this large amount of data. The exploration techniques presented in this work assist the physicians in making a diagnosis and scientists in comprehending the data.; Our model for the exploration of 3D volumetric data has two steps: the segmentation of the region of interest and its visualization. The segmentation and visualization techniques presented use a ray-based approach, where the data is analyzed and processed along a ray. We present three algorithms for segmentation that overcome the commonly encountered problem of partial volume effect. In one of the algorithms, we present segmentation rays that are special purpose rays designed to detect and optionally remove partial volume effect. Removing partial volume avoids obstruction of the region of interest when the data is visualized.; For visualizing segmented volumetric data we present a new paradigm of volumetric extraction. This technique can be used to either extract the region of interest or to remove unwanted regions from the data. Our extraction technique works in such a way that it maintains the smooth transition at object boundaries. This smooth transition allows high quality visualization of the region of interest using direct volume rendering. For rendering volumetric data, we present a new direct volume rendering technique that compensates for partial volume and avoids miss-classification and miss-coloring of partial volume voxels. Our technique presents a very simple transfer function interface to the user, avoiding the complexities of multi-dimensional transfer functions.; We further present a new paradigm in visualization that combines segmentation and classification during volume rendering. Unlike traditional volume rendering where an object's surface is displayed, our technique scans the object interior and displays classification information on the surface in different colors. We use a ray-based technique similar to the segmentation rays introduced earlier, to perform segmentation and classification on the object interior. We show the usefulness of our technique for the computer-aided detection of potential cancerous growth inside the colon. An important requirement for effective exploration of data is interactivity. We present an acceleration technique for direct volume rendering that uses ray-coherence to space leap rays when they traverse through transparent regions.