Region Based Segmentation and Classification of Multispectral Chromosome Images

  • Authors:

  • Petros S. Karvelis;Dimitrios I. Fotiadis;Alexandros Tzallas;Ioannis Georgiou

  • Affiliations:

  • Student member, IEEE/ University of Ioannina, Greece;Senior member, IEEE/ University of Ioannina, Greece;Student member, IEEE/ University of Ioannina, Greece;Medical School, Greece

  • Venue:
  • CBMS '07 Proceedings of the Twentieth IEEE International Symposium on Computer-Based Medical Systems
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Multiplex Fluorescent In Situ Hybridization (M-FISH) is a newly chromosome imaging technique where each chromosome class appears to have a distinct color. This technique although it makes the analysis of chromosome images easier, still exhibits misclassification errors that can be misinterpreted as chromosome abnormalities. A new method for the multichannel image segmentation and region classification is proposed. The segmentation of M-FISH images is based on a multichannel watershed segmentation method in order to define regions of same spectral characteristics. The region Bayes classification method which focuses on region classification is used. The classifier was trained and tested on nonoverlapping chromosome images and an overall accuracy 89% is achieved. The superiority of the proposed method over methods that use pixel-by-pixel classification is demonstrated.