Efficient topological cleaning for visual colon surface flattening

  • Authors:
  • Rui Shi;Wei Zeng;Jerome Zhengrong Liang;Xianfeng David Gu

  • Affiliations:
  • Department of Computer Science, Department of Radiology, Stony Brook University, Stony Brook, NY;School of Computing & Information Sciences, Florida International University, Miami, FL;Department of Computer Science, Department of Radiology, Stony Brook University, Stony Brook, NY;Department of Computer Science, Department of Radiology, Stony Brook University, Stony Brook, NY

  • Venue:
  • MICCAI'12 Proceedings of the 4th international conference on Abdominal Imaging: computational and clinical applications
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Conformal mapping provides a unique way to flatten the three dimensional (3D) anatomically-complicated colon wall. Visualizing the flattened 2D colon wall supplies an alternative means for the task of detecting abnormality as compared to the conventional endoscopic views. In addition to the visualization, the flattened colon wall carries supplementary geometry and texture information for computer aided detection of abnormality. It is hypothesized that utilizing both the original 3D and the flattened 2D colon walls shall improve the detection capacity of currently available computed tomography colonography. One of the major challenges for the conformal colon flattening is how to make the input colon wall inner surface to be genus zero, as this is required by the flatten algorithm and will guarantee high flatten quality. This paper describes an efficient topological cleaning algorithm for the conformal colon flattening pipeline. Starting from a segmented colon wall, the Marching Cube algorithm was first applied to generate the surface, then we apply our topological clearance algorithm to remove the topological outliers to guarantee the output surface is exactly genus 0. The cleared or denoised colon surface was then flattened by an Ricci flow. The pipeline was tested by 14 patient datasets with comparison to our previous work.