Three-dimensional reconstruction of blood vessel lumen from two X-ray projections
MetadataShow full item record
Local narrowing in arteries is caused by atherosclerotic plaques and often underlies ischemic stroke and heart attack. Vascular interventions for treating structural abnormalities require accurate measurement of lumen size and extent of narrowing. This work presents a novel method for three-dimensional (3D) reconstruction of blood vessel lumens deformed by atherosclerotic plaque from two x-ray projections. Most often measured from x-ray projection images, lumen sizes are poorly estimated when the lumen is deformed by atherosclerotic plaque. This is due in part to the dependence of apparent narrowing on the angle from which the projection image of the vessel is acquired. We have developed a model-based method for estimating lumen size and plaque from a single x-ray view independent of viewing angle. Cross-sectional diameters in healthy (undeformed) regions of the lumen are estimated using an elliptical model for cross-sectional shape; irregular cross-sectional shapes are represented as deviations from ellipticity. The deformed regions are automatically identified and plaque estimated using continuity of cross-sectional parameters along the vessel axis. Lumen size and percent stenosis measurements using x-ray angiograms of vessel phantoms yielded accurate results regardless of viewing angle. The algorithm was enhanced to identify and correct for the effect of branching and overlapping vessels on plaque estimation. Further evaluation using coronary angiograms yielded good results. A novel 'plaque image' has been developed as a two-dimensional tool for visualizing intruding plaque. The 3D lumen is reconstructed from two projections assuming that lumen cross sections are elliptical; deformations due to plaque are then generated using the plaque estimates from the individual projections. Reconstruction of the lumen of a vessel phantom using our technique (2 projections) compared well with a full reconstruction (180 projections) using a Feldkamp algorithm.