Image Processing and Analysis of Histopathological Images Relating to Hirschsprung’s Disease

  • Jinu Kurian Carleton University
  • Marco T.K. Law University of British Columbia
  • Victoria Madge McGill University
  • Adrian D.C. Chan Carleton University
  • Dina El Demellawy Children’s Hospital of Eastern Ontario
  • Eranga Ukwatta Carleton University

Abstract

The current procedure for surgical treatment of Hirschsprung’s disease involves histopathological imaging of excised colon post-surgery by an expert pathologist, to confirm the complete removal of diseased colon. Pathologists examine slices of colon for the presence of neurons (ganglions) which may innervate the intestinal muscle. However, this practice is time-consuming and subjective, with evaluations varying between experts. The percentage of HD patients with pathology indications, whose symptoms persist post-operation, encourage experts to find an objective measure for the improvement in surgical outcome. In this preliminary study with ten patient cases from the Children’s Hospital of Eastern Ontario, we are proposing an image processing pipeline to segment the muscularis propria and myenteric plexus regions, as initial steps to identifying ganglions. We were able to segment the muscularis propria using a unsupervised k-means clustering algorithm with an average dice coefficient of 71.22% ± 20.44%. Digital Image Subtraction Blue Enhancement (DISBE) was used to identify myenteric plexus regions with a precision of 70.53% ± 28.08% when using the manual segmentations for the muscularis propria. Promising results encourage further development of these algorithms
Published
2018-05-08
How to Cite
[1]
J. Kurian, M. Law, V. Madge, A. Chan, D. Demellawy, and E. Ukwatta, “Image Processing and Analysis of Histopathological Images Relating to Hirschsprung’s Disease”, CMBES, vol. 41, May 2018.
Section
Academic