Coronary Vessel Segmentation Algorithms for Cine-Angiography Image Data

Research output: Types of ThesesMaster's Thesis / Diploma Thesis

Abstract

In progress of a clinical study the General Hospital Linz (AKH Linz) elucidates in cooperation with the blood center of the Red Cross Linz the potential and impacts of stem cells for treatment of myocardial infarction patients. Adult autologous stem cells are expected to be able to build up any human tissue. Therefore, those unspecific cells are directly transplanted into the left ventricle to regenerate myocardial muscle areas affected by a heart attack. A software system developed at the FH OÖ F&E Competence Center Hagenberg (research and development department of the University of Applied Sciences Upper Austria) for diagnosis and monitoring of cardiac stem cell therapy is supporting the research group at AKH Linz. A dynamic three dimensional model of the human heart is generated from two dimensional biplane images acquired during x-ray angiography. Areas with reduced contractility are color-coded generating a much more expressive model than two dimensional grayscale images. Providing cardiologists a detailed view in 3D on the patient’s cardiac cycle. Next to the visualization of heart wall dynamics, the generated model also allows assessment of medical parameter like ejection fraction and wall motion patterns that gives a good picture about how efficient the heart is still working. The aim of this thesis is the development of novel manual and automated methods for segmentation of the coronary vessel tree in cine-angiograms. These segmented structures serve as input for three dimensional modelling and analysis. Developed algorithms are based on the application of a Gabor filter for contrast enhancement of vessel borders. In further steps statistical region classification contour detection of the vessel tree is done. Data is evaluated by concepts of mathematical morphology to provide automated methods and for centerline and border detection. The novel data model enables both automated and manual segmentation. Segmented regions are further classified corrected to achieve nearly optimal vessel preparation. Focus of this work is the development and implementation of novel algorithms for automated segmentation. Segmentation algorithms showed accurate results for coronary vessel segmentation.
Translated title of the contributionCoronary Vessel Segmentation Algorithms for Cine-Angiography Image Data
Original languageGerman
Publication statusPublished - 2005

Fingerprint Dive into the research topics of 'Coronary Vessel Segmentation Algorithms for Cine-Angiography Image Data'. Together they form a unique fingerprint.

Cite this