Towards 3-D LV shape recovery in biplane X-ray angiography using statistical shape models

Abstract

Coronary X-ray angiography has proven to be an efficient method for treatment and diagnosis of cardiovascular diseases. In clinical practice, quantitative LV analysis is done in 2-D and based on contour data since 3-D information is not available due to projection. In this work, a novel approach for recovering the 3-D LV shape from bi-planar X-ray images is presented. The sparse and noisy data available for reconstruction necessitates the incorporation of geometric prior information. A statistical shape model of the ventricular anatomy is learned from high-resolution multi-slice CT data. Reconstruction is based on a non-rigid 2-D/3-D registration technique. 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 simulated and real projection images. The presented method is evaluated using simulated data.