Global ventricular function was evaluated by both multiple gated cardiac blood pool scans (MUGA) and contrast ventriculograms in a group of 17 patients with suspected coronary artery disease. The contrast ventriculograms were analyzed frame by frame to generate a volume versus time curve for each patient, while the tracer data were analyzed by two methods: 1) the standard method, in which the left ventricle is identified on the end-diastolic frame and the background corrected activity under the region of interest obtained from the entire cardiac cycle, and displayed as a time versus activity curve; and 2) by a semi-automatic method in which the computer applies a threshold detection program to define the ventricular borders, and activity in the chamber at each point in the cardiac cycle is defined after background correction. The tracer data in each patient were analyzed independently by four observers. The tracer data correlated with the contrast data on a point by point basis r = 0.87 for the standard method, and 0.93 for the semi-automatic technique. An F test of variance revealed the semi-automatic method superior to the standard approach (P less than 0.05).