Intravascular ultrasound images vessel characterization using AdaBoost

Oriol Pujol, Misael Rosales, Petia Radeva, Eduard Nofrerias-Fernández

    Research output: Contribution to journalArticleResearchpeer-review

    25 Citations (Scopus)

    Abstract

    This paper presents a method for accurate location of the vessel borders based on boosting of classifiers and feature selection. Intravascular Ultrasound Images (IVUS) are an excellent tool for direct visualization of vascular pathologies and evaluation of the lumen and plaque in coronary arteries. Nowadays, the most common methods to separate the tissue from the lumen are based on gray levels providing non-satisfactory segmentations. In this paper, we propose and analyze a new approach to separate tissue from lumen based on an ensemble method for classification and feature selection. We perform a supervised learning of local texture patterns of the plaque and lumen regions and build a large feature space using different texture extractors. A classifier is constructed by selecting a small number of important features using AdaBoost. Feature selection is achieved by a modification of the AdaBoost. A snake is set to deform to achieve continuity on the classified image. Different tests on medical images show the advantages. © Springer-Verlag Berlin Heidelberg 2003.
    Original languageEnglish
    Pages (from-to)242-251
    JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume2674
    Publication statusPublished - 1 Dec 2003

    Fingerprint Dive into the research topics of 'Intravascular ultrasound images vessel characterization using AdaBoost'. Together they form a unique fingerprint.

    Cite this