Towards 3-D LV Shape Recovery in Biplane X-Ray Angiography Using Statistical Shape Models

Recovering the three-dimensional left ventricular shape from contrast-enhanced bi-planar x-ray image sequences is a challenging task. The inherently sparse projection data available for reconstruction and the ill-posed nature of the problem necessitate the incorporation of geometric prior knowledge. The novel approach presented in this work aims at reconstructing the endocardial surface from two projections using a statistical shape model that is learned from high-resolution multi-slice CT data. The three-dimensional statistical shape model of the left ventricle is built from 15 data sets acquired with a 64-slice CT. Corresponding points between the manually segmented endocardial surfaces are obtained by a novel automatic landmark generation approach. While other shape models often cut the left ventricle at the apical and basal area, our model retains anatomical details like the atrial concavity, the aortic valve region and the apex. In order to fit the shape model to the x-ray images of the patient, simulated projections of the model are calculated. An optimization procedure minimizes the difference between the silhouettes in the simulated x-ray images and the silhouettes in the recorded x-ray images. The presented method is evaluated based on simulated data. Ventricular shapes are reconstructed from simulated x-ray projections that are derived from the available CT images where the true three-dimensional form is visible. The application of statistical shape models yields anatomically plausible shapes. Including the atrial concavity, the aortic valve region and the apex in our model is necessary for deriving a complete ventricular silhouette and is relevant if densitometric information is incorporated for fitting the model. These anatomical areas usually overlap in the projections and are therefore hard to distinguish from the ventricular cavity without the presence of a priori information. A recovered shape is a potential base for further clinical evaluations.