Computer aided measurement of sub-epithelial collagen band in colon biopsies for collagenous colitis diagnosis

Vahid Malekian*, Rasoul Amirfattahi, Saeid Sadri, Mojgan Mokhtari, Alireza Aghaie, Mahdie Rezaeian

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

3 Citations (Scopus)


This paper presents a novel computer aided technique for screening of Collagenous Colitis (CC). CC is a type of microscopic colitis mostly characterized by chronic watery diarrhea which is a common feature with a range of other etiologies. Routine paraclinical tests from CC patients such as endoscopic and radiographic studies are usually normal, and diagnosis must be made by biopsy. The gold standard for a confirmative diagnosis of CC is to measure the thickness of the sub-epithelial collagen (SEC) in colon tissue samples. Visual inspection of microscopic samples is often time-consuming, cumbersome and subject to human errors. This fact demonstrates the necessity of developing an automated method which assists pathologists in evaluating histopathological samples more accurately in the busy clinical environment. To the best of our knowledge, this is the first time that a computer-assisted diagnosis algorithm has been applied to CC detection. The proposed method uses a pre-trained Multi-Layer Perceptron neural network to segment SEC band in colon tissue images. We compared a variety of different color and texture descriptors and explore the best set of features for this task. The investigation of the proposed method shows 94.5% specificity and 95.6% sensitivity rate.

Original languageEnglish
Pages (from-to)59-67
Number of pages9
Publication statusPublished - Feb 2013
Externally publishedYes


  • Adaboost
  • Collagenous colitis
  • Computer-assisted diagnosis
  • MLP neural network
  • Sub-epithelial collagen
  • Wavelet decomposition

Fingerprint Dive into the research topics of 'Computer aided measurement of sub-epithelial collagen band in colon biopsies for collagenous colitis diagnosis'. Together they form a unique fingerprint.

Cite this