Automated, nonrigid alignment of clinical myocardial contrast echocardiography image sequences: comparison with manual alignment

Abstract

Analysis of myocardial contrast echocardiography (MCE) images is currently done by manual techniques. The development of computationally efficient methods for aligning images provides an important first step toward the automation of MCE analysis. This is challenging because a nonrigid transformation correction is required. In this paper, we evaluate a state-of-the-art nonrigid alignment method on clinical MCE image sequences ( n = 58) acquired on patients during rest and dipyridamole stress, using both B-mode intermittent ultraharmonic (IUH) imaging and real-time myocardial perfusion imaging (RTMPI). Using manual alignment as the reference, we show quantitatively that the automated method aligns images as well as a human observer. However, the new method is faster and more reliable than manual alignment and removes the need for an experienced physician to perform it. The automated technique can be used for quick poststudy off-line analysis and has the potential to be incorporated into an ultrasound machine (E-mail: noble@robots.ox.ac.uk).