Abstract
Recovering the three-dimensional left ventricular shape from bi-planar contrast-enhanced x-ray images is a challenging task. The inhomogeneous saturation of contrast agent within the ventricle impairs the densitometric information often utilized for reconstruction. This and the inherently sparse projection data necessitate the incorporation of geometric prior knowledge. A novel approach is presented for obtaining a priori information of the left ventricle suitable for reconstruction. For the first time, a statistical shape model is built from high-resolution multi-slice computed tomography data for this purpose. While other shape models often cut the LV in the basal and apical area, our model retains anatomical details like atrial concavity, aortic valve region and apex. These areas are usually overlapped in the projections and therefore hard to distinguish from the ventricular cavity during reconstruction. Point correspondence between training shapes is determined by a new automatic landmark generation approach.
Originalsprache | Englisch |
---|---|
Titel | Challenges in Biosciences: Image Analysis and Pattern Recognition Aspects |
Herausgeber (Verlag) | Druckerei Riegelnick |
Seiten | 53-62 |
ISBN (Print) | 978-3-85403-232-8 |
Publikationsstatus | Veröffentlicht - 2008 |
Veranstaltung | 32nd Workshop of the Austrian Association for Pattern Recognition - St. Magdalena, Linz, Austria, Österreich Dauer: 26 Mai 2008 → 27 Mai 2008 |
Workshop
Workshop | 32nd Workshop of the Austrian Association for Pattern Recognition |
---|---|
Land/Gebiet | Österreich |
Ort | St. Magdalena, Linz, Austria |
Zeitraum | 26.05.2008 → 27.05.2008 |
Schlagwörter
- Heart
- Left Ventricle
- Biplanar X-Ray Angiography
- Computed Tomography
- Reconstruction
- Image Processing