Automatic nonrigid motion correction for quantitative first-pass cardiac MR perfusion imaging

First-pass dynamic contrast-enhanced cardiac magnetic resonance imaging is an increasingly important diagnostic tool for coronary artery disease. It is typically performed with breath holding and electrocardiogram gating to minimize motion, but movement caused by residual respiration, cardiac arrhythmia, or missed gating triggers during image acquisition will induce nonrigid deformations of the myocardium that can hamper interpretation and perfusion quantification. We propose an automatic nonrigid image registration framework to correct motion in a series of cardiac magnetic resonance perfusion images. Our method employs reference frame detection combined with robust flow field estimation using a large displacement optical flow formulation coupled with post-hoc image warping and flow field distortion correction. This framework is multi-threadable and can be applied to standard myocardial series, arterial input function series, and proton density weighted images to facilitate perfusion quantification.