BACKGROUND: -Accurate measures are critical when attempting to distinguish normal from pathological changes in cardiac function during exercise, yet imaging modalities have seldom been assessed against invasive exercise standards. We sought to validate a novel method of biventricular volume quantification by cardiac magnetic resonance imaging (CMR) during maximal exercise. METHODS AND RESULTS: -CMR was performed on 34 subjects during exercise and free-breathing using an ungated real-time CMR ("RT-ungated") sequence. ECG and respiratory movements were retrospectively synchronized enabling compensation for cardiac cycle and respiratory phase. Feasibility of RT-ungated imaging was compared with standard exercise CMR imaging with ECG gating ("gated"), Accuracy of RT-ungated CMR was assessed against an invasive standard (direct Fick) and reproducibility was determined following a second bout of maximal exercise. Ventricular volumes were able to be analyzed more frequently during high-intensity exercise using RT-ungated as compared with gated CMR (100% vs. 47%, p<0.0001) and with better inter-observer variability for RT-ungated (coefficient of variation CV=1.9% and 2.0% for left and right ventricular stroke volumes, respectively) than gated (CV=15.2% and 13.6%), p <0.01. Cardiac output determined by RT-ungated CMR proved accurate against the direct Fick method with excellent agreement (intraclass correlation coefficient R=0.96) which was highly reproducible during a second bout of maximal exercise (R=0.98). CONCLUSIONS: -By combining real-time ungated CMR with post-hoc analysis incorporating compensation for respiratory motion, highly reproducible and accurate biventricular volumes can be measured during maximal exercise.