To compare radionuclide end-diastolic (EDV) and end-systolic (ESV) volumes with angiographic volume, we studied 52 patients with equilibrium radionuclide angiography using 99mTc-human serum albumin within 48 hours of contrast angiography. Each RR interval was divided into 20--28 equally timed frames and a time-activity curve generated. End-diastolic counts were taken at the early peak of the curve and end-systolic counts at its nadir. Counts were divided by the total number of processed heart beats and normalized for: 1) dose per body surface area; 2) plasma volume; and 3) counts/ml of plasma. A cardiac phantom was developed and serial volumes were studied using a normalization factor. Radionuclide values were expressed as dimensionless units and compared with either biplane angiographic volumes (in the patient studies) or known phantom volumes. Good correlations were obtained with methods 1 and 2 in 35 patients (r greater than 0.84), but the best correlation was obtained in 17 patients when normalization for counts/ml of plasma was used (r = 0.98; y = 0.255 x -0.121). The standard error of the estimate (SEE) was +/- 11.5 ml for EDV and +/- 7.3 ml for ESV. The phantom study also showed an excellent correlation (r = 0.99), with a SEE of +/- 6.5 ml. We conclude that a radionuclide method independent of geometric assumptions can be used to estimate left ventricular volume in man.