The commonly recommended one-dimensional (1D) and two-dimensional (2D) algorithms for left ventricular (LV) mass calculation are limited by assumptions about ventricular geometry and image plane position. To assess the accuracy of these algorithms after eliminating errors associated with image plane position, LV mass was calculated from high quality cardiovascular magnetic resonance imaging (CMR) data sets using ID (modified cube formula; MCF) and 2D algorithms [area-length (AL) and truncated ellipsoid (TE) methods], and the summation of slices (SS) method as reference technique in 25 patients with LV aneurysms, 15 patients with hypertrophic cardiomyopathy, and 10 healthy subjects. Each algorithm in each group overestimated LV mass compared to SS (p < 0.05 and p < 0.001). In each patient group, the smallest bias to the reference method was observed for the TE algorithm (p < 0.001 vs. MCF and p < 0.05 vs. AL). The LV mass interval encompassing the limits of agreement was 120-220 g for MCF, 100-148 g for AL, and 80-136 g for TE. The interstudy reproducibility of the SS technique for the assessment of LV mass was superior compared to the 1D and 2D algorithms. We conclude that despite the use of optimized image plane position 1D and 2D algorithms are inaccurate for calculation of LV mass in ventricles with normal and distorted LV geometry. Thus, 3D imaging techniques, such as CMR, should be preferred when assessing LV mass.