Medical Community: Abstracts




Sarang Lakare, Dongqing Chen, Lihong Li, Arie Kaufman, Mark Wax, and
Zhengrong Liang

Purpose: Virtual colonoscopy aims to provide a safe and comfortable technique to examine the colon for polyps (cancerous growth). In theory, an accurate diagnosis requires complete coverage of 100% of the lumen surface. Mori et al. presented a method to track observed regions and present unobserved areas to the user (SPIE 4321 pp 134-145). We have developed a similar method that automatically displays, places in sized order and allows viewing of the areas of the colon surface not visualized during initial navigational viewing. While complete surface visualization is possible, we demonstrate that all of these "missed patches" do not have to be reviewed to detect clinically significant colon polyps.

Methods: Breath-hold, supine and prone spiral CT scans were performed on 103 patients after bowel preparation and colon distention with 2L of CO2. After automatic segmentation and electronic cleansing of the colon lumen, the medial axis (centerline) was extracted. Volume rendering fly-through was performed and visualized surfaces were marked taking into account what Mori termed the "physical factor" and the "psychological factor". Instead of only using Mori's description of a marked single voxel per ray, specifically the one with a opacity above a certain threshold, our system marks all voxels once the opacity passes a minimum threshold until it passes the maximum threshold. This more closely resembles how voxels contribute to the image in volume rendering. Also, we incorporate a "psychological factor" that allows including voxels within 6 centimeters of the viewpoint to be classified as visualized. In this application, colon surfaces far away from the virtual viewpoint will not be counted since the user cannot adequately analyze those areas. The virtual camera was passed both antegrade and retrograde, and the combined visible surface voxel count was recorded. After both fly-throughs, all "patches" of connected surface area not yet seen were identified, measured, sorted by size, and counted. Clinically significant patches * defined as smallest diameter being 35mm * were sequentially visualized by stepping through the sorted list, followed by all patches. The total visualized surface was recorded for each.

Results: Virtual fly through in one direction viewed an average of 80% (range 61-91) of the colon surface, primarily missing the backsides of haustral folds and around sharp bends. Combined antegrade and retrograde flythroughs viewed an average of 94% (range 84-99) of the surface. Even with 94% average coverage, there was an average of 40 (range 13-91) clinically significant patches not yet visualized. After identifying and viewing these areas, 98-100% of the lumen surface was seen. Even if all patches, regardless of size, could be viewed to achieve 100% surface coverage, this extra effort would be unnecessary to detect clinically significant, larger than 5mm, polyps.

New or breakthrough work: Compared to previous approaches, this work more accurately marks visualized surfaces when performing volume rendering based navigation. It also takes into account psychological factors of the viewer to not include areas far away from the viewpoint. A new method of automatically identifying and visualizing the "missed patches" that sorts them according to size allows the reviewer to step through the "clinically significant" patches, thus saving interpretation time.

Conclusion: By enabling endoscopic navigation along the colon centerline, in both antegrade and retrograde directions, virtual colonoscopy allows evaluation behind haustral folds and around sharp bends, thereby visualizing significantly more surface area than optical colonoscopy. However, this still leaves a considerable number of areas unseen. Automatically marking the visualized surface area and identifying and viewing unseen patches allow examination of all clinically significant surfaces of the colon.

Keywords: Virtual Endosocopy, Volume Rendering, Interactive
Navigation, Complete Surface Coverage

Principal Author Affiliation:
State University of New York at Stony Brook
62 Washington St.
East Setauket, NY 11733
Phone: (631) 444-6576
Fax: (631) 444-9701

Principal Author Biography:
Kevin is a Ph.D. candidate in Computer Science at SUNY
Stony Brook. His research interests include
Hardware Architectures for Volume Visualization and Medical Imaging

Back to Abstracts