Segmenting high-frequency intracardiac ultrasound images of myocardium into infarcted, ischemic, and normal regions

Research output: Contribution to journalArticle

22 Scopus citations

Abstract

Segmenting abnormal from normal myocardium using high-frequency intracardiac echocardiography (ICE) images presents new challenges for image processing. Gray-level intensity and texture features of ICE images of myocardium with the same structural/perfusion properties differ. This significant limitation conflicts with the fundamental assumption on which existing segmentation techniques are based. This paper describes a new seeded region growing method to overcome the limitations of the existing segmentation techniques. Three criteria are used for region growing control: 1) Each pixel is merged into the globally closest region in the multifeature space. 2) "Geographic similarity" is introduced to overcome the problem that myocardial tissue, despite having the same property (i.e., perfusion status), may be segmented into several different regions using existing segmentation methods. 3) "Equal opportunity competence" criterion is employed making results independent of processing order. This novel segmentation method is applied to in vivo intracardiac ultrasound images using pathology as the reference method for the ground truth. The corresponding results demonstrate that this method is reliable and effective.

Original languageEnglish (US)
Pages (from-to)1373-1383
Number of pages11
JournalIEEE transactions on medical imaging
Volume20
Issue number12
DOIs
StatePublished - Dec 1 2001

Keywords

  • Equal opportunity competence
  • Geographic similarity
  • High-frequency intracardiac echocardiography image
  • Seeded region growing method

ASJC Scopus subject areas

  • Software
  • Radiological and Ultrasound Technology
  • Computer Science Applications
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Segmenting high-frequency intracardiac ultrasound images of myocardium into infarcted, ischemic, and normal regions'. Together they form a unique fingerprint.

  • Cite this