 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Highlight
|
Specializing Interfaces
User Experience and Interface Research |
|
| Specializing Interfaces for Segmentation of Time-Varying Volumetric Data: Within volumetric data, the visual characteristics of the data itself (graininess, lack of natural light source or depth cue information, black and white versus real-world color spectrum, lack of familiarity with the structures represented, the similarity inherent in biomorphic shapes etc.) do not provide the visual or depth cues that human beings routinely utilize in navigating spaces or interacting with objects in the real world. These challenges are compounded by the fact that as bio-imaging data increases in resolution it normally decreases in extent, producing both visual fragmentation and spatial disorientation for users interacting with data at multiple scales and resolutions. While automatic methods for segmentation are becoming more reliable, many data sets still require manual intervention, either because of poor image quality or the need for an expert to review and edit results from an automatic technique. Additionally, navigating through volumetric biological data and retaining a felt-sense of one’s spatial orientation within the volume is a significant difficulty in the segmentation process. As a result, viewing and manually segmenting volumetric data is a challenging task even for bio-imaging experts, positioning segmentation as a bottleneck in medical and scientific endeavors. Our aim is to accelerate discovery by streamlining the segmentation process and making it possible for experts and non-experts to accomplish effective and accurate segmentations. In developing methods for segmentation of time-varying volumetric data, our questions include: How do we display 3D data in a 2D environment so that users can understand where they are within the volume? What enables users to understand navigation within a volume when utilizing oblique views rather than parallel fixed orientation slices through the volume? How can we best represent 3D data within a 2D environments routinely used for segmentation so that users can identify features and grasp what they are viewing? How can we “teach” novice users what it means to segment data, and what a contour and a set of contours actually portrays? And, How do we give users an understanding of the quality and completeness of their contributions? The understanding of these concepts is key to producing meaningful and usable results from complex, time-varying volumetric data. This NSF sponsored research (DEB1053566) brings together researchers from UCSD (Ruth West, PI) and Washington University St. Louis (Cindy Grimm, PI). This project is currently in development. |
|
 |
|
An interdisciplinary artist-researcher working with digital media and interactive technologies.
