GRETSI'03 19e Colloque GRETSI
sur le traitement du signal et des images
Paris   8 - 11 septembre 2003
Accueil Programme Par session Par auteur Par thème Par code

Informations concernant l'article

Titre
Classification of hepatic metastasis in enhanced CT images by dipolar decision tree
Auteur(s)
Marek Kretowski Bialystok Technical University
Johanne Bezy-Wendling INSERM, Universite de Rennes 1
Dorota Duda Bialystok Technical University
Références
vol. III, page 327
 L'article au format PDF
 
Pour obtenir Acrobat Reader (version 5 minimum recommandée) nécessaire pour sa lecture.

Résumé

In the paper, computer-aided decision support system for hepatic metastasis diagnosis based on dynamic CT scans is presented. Analyzed image database contains 4 types of pathological lesions and normal tissue in typical acquisition time moments of dynamic CT of the liver: without contrast material and after injection, in arterial and portal phases. In the proposed approach, texture analysis of drawn ROI-s is combined with a new classification method - dipolar decision tree. In this method, each node of the binary tree corresponds to a multivariate test (hyper-plane). Searching for an optimal position of the hyper-plane is based on separation of mixed dipoles (pairs of input feature vectors from different classes). Experimental results show that the proposed approach allowed obtaining the competitive classification quality. Furthermore, it could be improved by taking into account the acquisition time of CT images.
Edition : Télécom-Paris -- 2003