Count-based scintigraphic left ventricular end-diastolic (LVED) volume measurement was optimized using a reproducible method for determining left ventricular counts and an independently measured average apparent tissue attenuation coefficient (0.16 cm-1). Tissue depth was calculated by triangulation. Results were compared to single-plane contrast ventriculographic volumes by an area-length method, performed within one hour, in 18 patients. The overall correlation of measurements of LVED volume by the 2 methods was 0.96 with standard error of the scintigraphic estimate of 15.8 ml. For 6 patients with angiographically normal wall motion, the correlation of volume measurements was 0.99 with standard error of the estimate of 5.1 ml. The mean absolute difference in LVED volume by the 2 methods was 3.8 ml in the group with normal wall motion compared to 19.2 ml in the 12 patients with angiographically abnormal wall motion. Area-length LVED volume calculation assumes that the left ventricle conforms to a standard shape. Discrepancies in volume estimates with abnormal ventricular wall motion suggest that the area-length method is less accurate. Optimized count-based LVED left ventricular volume measurement is accurate and might be preferable to single-plane contrast angiographic volume measurement of abnormal ventricles.