This study compared the accuracy and reproducibility of three previously described and one new radionuclide method of measuring left ventricular volumes in 19 subjects using contrast ventriculographic volumes (n = 38, mean volume = 126.6 ml) as the gold standard. The four methods were compared using both manual and automated ROIs. For manual ROIs, the Links (189.7 ml, r = 0.85), Starling (183.2 ml, r = 0.77) and the new count ratio method (141.4 ml, r = 0.90) overestimated contrast volumes, while the Massardo method (122.5 ml, r = 0.91) provided accurate volumes. For the automated ROIs, we performed an interpolative background subtraction and used a 50% threshold of the highest count pixel to define the ventricular regions. The automated Massardo method severely underestimated the contrast volume (59.5 ml, r = 0.90), while the other automated methods yielded accurate volumes: Links (122.4 ml, r = 0.89), Starling (118.1 ml, r = 0.81) and the new count ratio method (125.0 ml, r = 0.90). The interobserver reproducibility of the automated methods was excellent (mean difference = 1%-4%) compared to the manual methods (2%-8%). Because no additional images, blood counting, attenuation, or decay correction were necessary, the manual Massardo method and the automated count ratio method are the simplest to perform. We conclude that automated determination of left ventricular volumes using the new count ratio method is rapid, accurate, reproducible and could readily be incorporated into routine clinical use.