Rigid Motion Estimation using Accelerated Iterative Coordinate Descent (REACT) for MR Imaging

A team from Stanford University has published a new paper on arXiv introducing REACT (Rigid Motion Estimation using Accelerated Iterative Coordinate Descent), a computational method designed to estimate and correct 3D rigid motion in Magnetic Resonance Imaging. The core innovation is an autofocus technique that assumes a piecewise-constant motion trajectory. It breaks down the complex, high-dimensional optimization problem of estimating motion parameters for individual temporal segments into a series of more manageable subproblems, each solved using a derivative-free solver. This approach avoids the computational burden of an exhaustive grid search.

The method's feasibility was demonstrated through numerical simulations and in vivo evaluation. Researchers tested REACT on free-breathing coronary MR angiography datasets acquired using a 3D cones trajectory with image-based navigators. They quantified coronary artery sharpness using a metric called unbounded image edge profile acutance (u-IEPA). Results showed REACT outperformed a conventional translational motion-estimation method, yielding higher sharpness for both the left anterior descending artery (LAD) and right coronary artery. Notably, it also produced higher u-IEPA for the LAD than a comparative autofocus nonrigid motion correction method, demonstrating its effectiveness for a specific, challenging clinical imaging scenario.