Rigid and elastic registration for coronary artery IVUS images.

Abstract

Intravascular ultrasound (IVUS) has been widely used in diagnosis and interventional treatment of cardiac vessel diseases. The coronary artery IVUS images are usually polluted by motion artifacts caused by cardiac motion, pulsatile blood and catheter twist during continuous pullback acquisition. Strategies for rigid and elastic registration of coronary artery IVUS studies are developed to suppress the longitudinal motion and misalignment between successive frames. Rigid registration is performed by searching for the optimal matching for each frame in other cycles based on the cyclic variation of gray-scale features. The image sequence is gated to properly identify the frames in each cardiac phase. Then, elastic registration between frames is achieved through an optimization algorithm based on thin plate spline (TPS) to correct the misalignment of successive slices. Experimental results with in vivo image data shows that the rigid registration performs better than the offline ECG gating. The elastic mapping relation between lumen contours in successive frames is smooth and continuous. The serrated vessel wall borders in longitudinal cuts are smoothed after rigid registration while image segmentation and feature extraction are required. The point-to-point correspondence between lumen contours detected from two matched frames is obtained with elastic registration.