The purpose of this study was to test motion-correction algorithms for initial-transit radionuclide angiocardiograms acquired at rest and during bicycle and treadmill exercise. Treadmill data was spatially reoriented by computer software designed to eliminate motion of a 125I point source simultaneously recorded at a lower energy window. A second algorithm based on left ventricular centroid counts further corrected for motion on all studies. Exercise left ventricular ejection fraction was higher on the treadmill (0.68 +/- 0.07) compared to the bicycle (0.64 +/- 0.08) (p less than 0.0001, r = 0.88). Treadmill exercise also resulted in larger end-diastolic volumes (180 +/- 30 versus 157 +/- 36, p less than 0.0001), stroke volumes (124 +/- 28 versus 101 +/- 29, p less than 0.0001) and cardiac outputs (19.9 +/- 4.6 versus 15.9 +/- 5.0, p less than 0.0001). Similar variances for these hemodynamic measurements suggest that the mean differences observed were physiologic and that error from body motion was effectively corrected by this approach. We conclude that the measurement of left ventricular function during treadmill exercise, when combined with these techniques for correcting motion, is a reasonable alternative to conventional bicycle exercise.