Continuous Ventricular Volumetric Quantification in Patients with Arrhythmias using Real-Time 3D CMR-MOTUS

Conventional cardiovascular magnetic resonance (CMR) cine imaging relies on binning multiple heartbeats into a single cardiac cycle, which fails in arrhythmic patients where beat-to-beat variability causes motion artifacts and loss of functional information. Real-time 2D imaging captures individual beats but lacks volumetric coverage for mapping arrhythmic cardiac dynamics. We present a 3D real-time motion-field reconstruction method enabling continuous volumetric assessment in patients with premature ventricular contractions (PVCs) using a free-running CMR protocol. CMR-MOTUS was extended to jointly reconstruct real-time 3D motion fields and a motion-corrected reference image from continuous, ungated, non-breath-held data acquired with a variable-density Cartesian OPRA trajectory. Beat-to-beat ejection fraction (EF) was computed by propagating a single segmentation through all frames using the reconstructed motion fields. The method was validated on a cardiac motion phantom and tested in four healthy volunteers and four PVC patients. Phantom EF closely matched ground truth (22.1 +/- 0.6 percent vs. 21.9 percent). In healthy volunteers, EF values agreed with 2D references and showed narrow distributions reflecting physiological consistency. In PVC patients, EF distributions were bimodal, with the lower mode corresponding to PVC beats with markedly reduced EF. ECG confirmed alignment between EF irregularities and PVC episodes. These results show that 3D real-time motion-field reconstruction enables continuous beat-to-beat volumetric quantification in arrhythmia, revealing functional heterogeneity that conventional binning obscures. The bimodal EF distributions capture the true hemodynamic impact of PVCs and may provide clinically relevant metrics for monitoring and treatment evaluation.