Leiden University Scholarly Publications

Oost, C.R. (2008)

Model-driven segmentation of X-ray left ventricular angiograms

Doctoral Thesis

X-ray left ventricular (LV) angiography is an important imaging modality to assess cardiac function. Using a contrast fluid a 2D projection of the heart is obtained. In current clinical practice cardiac function is analyzed by drawing two contours manually: one in the end diastolic (ED) phase and one in the end systolic (ES) phase. From the contours the LV volumes in these phases are calculated and the patient__s ejection fraction is assessed. Drawing these contours manually is a cumbersome and time-consuming task for a medical doctor. Furthermore, manual drawing introduces inter- and intra-observer variabilities. The focus of the research presented in this thesis was to automate the process of contour drawing in X-ray LV angiography. The developed method is based on Active Appearance Models. These statistical models, in which the cardiac shape and the cardiac appearance are modeled, have proven to be able to mimic the drawing behavior of an expert cardiologist. The clinical...Show moreX-ray left ventricular (LV) angiography is an important imaging modality to assess cardiac function. Using a contrast fluid a 2D projection of the heart is obtained. In current clinical practice cardiac function is analyzed by drawing two contours manually: one in the end diastolic (ED) phase and one in the end systolic (ES) phase. From the contours the LV volumes in these phases are calculated and the patient__s ejection fraction is assessed. Drawing these contours manually is a cumbersome and time-consuming task for a medical doctor. Furthermore, manual drawing introduces inter- and intra-observer variabilities. The focus of the research presented in this thesis was to automate the process of contour drawing in X-ray LV angiography. The developed method is based on Active Appearance Models. These statistical models, in which the cardiac shape and the cardiac appearance are modeled, have proven to be able to mimic the drawing behavior of an expert cardiologist. The clinical parameters, as determined by the automated method, showed a similar degree of accuracy as when determined by an expert. Furthermore, the required time for patient analysis was reduced considerably and the inter- and intra-observer variabilities were structurally decreased.Show less

Active Appearance Models

Automated Delineation

X-ray Left Ventricular Angiography

All authors

Oost, C.R.

Supervisor

Reiber, J.H.C.

Co-supervisor

Lelieveldt, B.P.F.

Qualification

Doctor (dr.)

Awarding Institution

Leiden University Medical Center (LUMC), Leiden University

Date

2008-09-30

ISBN (print)

9789090233987