Background: A single breath-hold evaluation of ventricular function would allow assessment in cases where scan time or patient tolerance is limited. Spatiotemporal acceleration techniques such as TSENSE decrease cardiovascular MR acquisition time, but standard slice summation analysis requires enough short axis slices to cover the left ventricle (LV). By reducing the number of short axis slices, incorporating long axis slices, and applying a 3D model based analysis, it may be possible to obtain accurate LV mass and volumes. We evaluated LV volume, mass and ejection fraction at 3.0T using a 3D modeling analysis in 9 patients with a history of myocardial infarction and one healthy volunteer. Acquisition consisted of a standard short axis SSFP stack and a 15 heart-beat single breath-hold six slice multi-planar (4 short and 2 long axis) TSENSE SSFP protocol with an acceleration factor of R = 4. Results: Differences (standard minus accelerated protocol mean +/- s.d.) and coefficients of variation (s.d. of differences as a percentage of the average estimate) were 7.5 +/- 9.6 mL and 6% for end-diastolic volume (p = 0.035), 0.4 +/- 5.1 mL and 7% for end-systolic volume (p = NS), 7.1 +/- 8.1 mL and 9% for stroke volume (p = 0.022), 2.2 +/- 2.8% and 5% for ejection fraction (p = 0.035), and -7.1 +/- 6.2 g and 4% for LV mass (p = 0.005), respectively. Intra-and inter-observer errors were similar for both protocols (p = NS for all measures). Conclusion: These results suggest that clinically useful estimates of LV function can be obtained in a TSENSE accelerated single breath-hold reduced slice acquisition at 3T using 3D modeling analysis techniques.