VISUALIZATION: RUI: User-Directed Segmentation and Visualization of Volumetric Data Sets in a Collaborative, Remotely Hosted Environment

University Of Wisconsin-La Crosse, La Crosse WI

Investigators

Abstract

By their nature, volumetric data sets, such as CT, MRI and the Visible Human cryosection images, are not subject to the same form of visual inspection and cognitive refinement that human experts apply when viewing and interpreting 2D images. The internal cycle of choices made by humans as they visually survey an image is disrupted and forced into an external process where users must adapt to the technical requirements of the visualization system. This project proposes to continue research into an immersive visualization system that is a first step toward creating a user interaction paradigm that more naturally allows users to apply their interpretive expertise to the visualization and segmentation of volumetric data sets. In this environment, segmentation, the process of isolating discrete structures within a data volume, is performed in parallel with visualization. Users have continuous control of the visualization and segmentation algorithms and can freely exercise their expert interpretive knowledge. The current implementation, called the Immersive Segmentation Environment, emphasizes the controlled application of high computational cost algorithms to local regions of the data volume. Using stereo graphics, the user is presented with the illusion that the data volume occupies physical space. The user interacts with the system using a 3D space tracked probe. Visualization and segmentation algorithms continuously operate on the data volume within a local neighborhood of the user.s position. This neighborhood corresponds to the user.s visual focus point. The algorithms initialize themselves based upon an analysis of the local neighborhood and heuristics. The system is implemented as a client/server application and consequently allows a user at a low-cost workstation to effectively use remote computational resources to visualize large volumetric data sets. The system implements two main classes of segmentation algorithms. Both determine initial algorithm parameters based upon a cluster analysis of the local neighborhood and are characterized by an iterative refinement approach. The proposed research will continue to investigate the capabilities of this unique user interaction model for visualizing, exploring and interpreting volumetric data sets. The research plan contained in this proposal is a natural outgrowth of prior work performed by the PI. This ongoing work has had a positive impact on the curriculum, student research projects and faculty research within the Computer Science department at the UW-La Crosse.