Automatic malignancy classification of colon tumor by feature analysis of biopsy images

Takatoshi Karino, Toshiyuki Tanaka

Research output: Contribution to conferencePaperpeer-review

Abstract

We intend to construct a system to automatically diagnose malignant potential of colon biopsies. First, we classified biopsies components (nuclei, cytoplasm and interstitium) by discriminant analysis method with Mahalanobis's generalized distance and Otsu method twice. Furthermore, only ductal nuclei were sorted out from the other nuclei by using differences of areas and densities of edges, that is high in lymphoid follicles. We proposed a method for quantifying pseudostratified nuclei by a measure of thickness of ductal nuclei and methods for quantifying structural atypia by fractal dimension analysis with Box-counting method and HLAC features. Moreover, poorly and moderately differentiated carcinoma in Group5 was able to be discriminated with high accuracy by density of labeled ranges. With these features, biopsies' images were classified into 4 Groups by support vector machine.

Original languageEnglish
Pages1820-1824
Number of pages5
Publication statusPublished - 2013 Jan 1
Event2013 52nd Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2013 - Nagoya, Japan
Duration: 2013 Sep 142013 Sep 17

Other

Other2013 52nd Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2013
CountryJapan
CityNagoya
Period13/9/1413/9/17

Keywords

  • Biopsy test
  • Cancer
  • Colon
  • Image processing

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Computer Science Applications
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Automatic malignancy classification of colon tumor by feature analysis of biopsy images'. Together they form a unique fingerprint.

Cite this