A New Visualization Method for Virtual Colonoscopy

Virtual colonoscopy or `colonography' is a patient-friendly, modern screening technique for polyps. Automatic detection of polyps can serve to assist the radiologist. This paper presents a method based on clustering the principal curvatures. Via automatic polyp detection 5/6 polyps (>5 mm) were detected at the expense of 9 false positive findings per case. For visualization, the bowel surface is presented to the physician in a `panoramic' way as a sequence of unfolded cubes. Conventionally, only 93% of the colon surface is available for examination. In our approach the area in view is increased to 99.8%. The unfolded cube visualization is another step to optimize polyp detection by visual examination. Experiments show a sensitivity of 10/10 (on a per patient basis) for any polyp. The specificity was 7/10.

@InProceedings { VSGPTGSV01a,
  author       = "Vos, F.M. and Serlie, I.W.O. and Gelder, R.E. van and Post, Frits H. and Truyen, R. and Gerritsen, F.A. and
                  Stoker, J. and Vossepoel, A.M.",
  title        = "A New Visualization Method for Virtual Colonoscopy",
  booktitle    = "Medical Image Computing and Computer-Assisted Intervention - MICCAI 2001",
  month        = "October",
  year         = "2001",
  editor       = "Wiro J. Niessen and Max A. Viergever",
  publisher    = "Springer-Verlag",
  note         = "0302-9743",
  url          = "http://graphics.tudelft.nl/Publications-new/2001/VSGPTGSV01a"
}