Real-time 2D-3D MR cardiac image registration during respiration using extended Kalman filter predictors

Real-time cardiac image registration is advantageous in integrating real-time (RT) images with priory and complementary images of the myocardium. Myocardial stem cell delivery and radiofrequency ablation are some cases that could benefit RT registration. Most of these applications, However, take long time and should get along with respiratory motion. On the other hand, registration is not that sharp to compensate this motion. Time series prediction techniques could compensate this shortcoming by estimating the heart displacements caused by respiratory motion. In this study we proposed a three-stage framework for RT 2D-3D cardiac image registration during respiration. The achieved mean misalignment through three complex simulations was less than 2 mm, which is a clinically acceptable threshold.