Non-linear image registration for correction of motion artifacts during optical imaging of human hearts

Optical imaging of cardiac electrical activity can be used to elucidate patho-physiological mechanisms of cardiac arrhythmias. However, cardiac motion during optical imaging causes significant error in electrophysiological measurements such as action potential duration. In particular, cardiac tissue in fibrillation introduces highly non-linear imaging artifacts. We present a novel approach that uses non-linear image registration to correct for in-plane cardiac motion, particularly of non-linear origin found during cardiac arrhythmias. The algorithm is performed entirely post-acquisition and does not require a complicated optical setup. It is computationally fast, and available as open source. The algorithm was tested with images acquired from five excised dilated myopathic human hearts and the results show that the image registration method significantly reduces both non-linear and linear motion-related artifacts in both sinus rhythm and ventricular fibrillation. This algorithm corrects for non-linear imaging artifacts caused by cardiac motion that are impossible to correct using linear registration methods.