Résumé : Background. Two different algorithms, which are fast and automatic and which operate in 3-dimensional space, were compared in the same group of patients to compute left ventricular ejection fraction (LVEF) and volumes from gated blood pool tomography. One method, developed at Cedars-Sinai Medical Center (CS), was dependent on surface detection, whereas the other method, developed at the Free University of Brussels (UB), used image segmentation. Methods and Results. Gated blood pool tomograms were acquired in 92 consecutive patients after injection of 740 MBq of technetium 99m-labeled human serum albumin. After reconstruction and reorientation according to the left ventricular long axis, LVEF and left ventricular volumes were measured with the CS and UB algorithms. Measurements of LVEF were validated against planar radionuclide angiocardiography (PRNA) results. The success rates of the algorithms were 87% for CS and 97% for UB. Agreement between LVEF measured with CS and UB (LVEFCS = 0.91 · LVEFUB - 0.85; r = 0.87) and between LVEF measured with CS and PRNA (LVEFCS = 1.04 · LVEFPRNA - 4.75; r = 0.80) and UB and PRNA (LVEFUB = 0.98 · LVEFPRNA + 4.42; r = 0.82) was good. For left ventricular volumes, linear regression analysis showed good correlation between both methods with regard to end-diastolic volumes (r = 0.81) and end-systolic volumes (r = 0.91). On average, end-diastolic volumes were similar and end-systolic volumes were slightly higher with CS than with UB. Consequently, significantly lower LVEFs were observed with CS than with UB. Conclusions. Good correlation was observed between CS and UB for both left ventricular volumes and ejection fraction. In addition, measurements of LVEF obtained with both algorithms correlated fairly well with those obtained from conventional PRNA over a wide range of values.