Machine-Learned Assist 
for Boundary Contour Tracing

Presented at the Second Visible Human Project Conference, October 1998
 
Stewart Crawford-Hines & Charles Anderson 
sgcraw@cs.colostate.edu  & anderson@cs.colostate.edu 

Department of Computer Science 
Colorado State University 
Fort Collins, CO 80523 

 Thomas McCracken 
vip@visiblep.com 

Visible Productions 
116 North College Ave, Suite #7 
Fort Collins, CO 80523
 
ABSTRACT
To generate anatomical models from the Visible Human imagery, a key first step is to trace the tissues & organs of interest on each of many slices. This tracing has been, by and large, a manual task so far.  Our focus is to aid the human tracers by having a computer system learn the contour being traced, and then automatically trace through similar regions. We have integrated our research ideas into the production tracing systems of Visible Productions, and early results demonstrate a significant improvement in overall tracing speed, while potentially relieving human tracers of much of the monotony in their task.
IMAGE LIST

OUTLINE OF PAPER

  1. INTRODUCTION
  2. LEARNING & PROJECTING A TRACE
  3. SUMMARY RESULTS
  4. CURRENT DIRECTIONS
  5. REFERENCES
 

ACKNOWLEDGMENTS

This research is funded in part by CASI (Colorado Advanced Software Institute) and NSF (SBIR#9761590).  Our research software was integrated into the tracing software of Visible Productions, Inc. (of Fort Collins, CO); Dan Steward provided invaluable system assistance.  And, of course, thanks to NIH for making the Visible Human imagery widely available.