We investigated the frequency and extent of changes in heart position and geometry independent of body motion during stress single-photon emission tomography (SPET) thallium-201 myocardial perfusion imaging. Following an exercise treadmill test, patients had a 22.1-min SPET acquisition which was followed immediately by a static image acquisition for 1 min with the camera position identical to the first view of the SPET study. Point sources were placed on the body to monitor patient motion. Cardiac motion was assessed by an approach which mimicked a cross-correlation technique applied to cardiac count profiles along the horizontal and vertical directions from the first view of the SPET study and the static image. A large percentage (87.5%) of cases had some degree of horizontal or vertical motion. Pixel shifts in cardiac position of > or = 2 pixels (12 mm) occurred in 60% of patients. In 37% of patients who moved the cardiac motion was consistent with simple translation of the heart and thus amenable to correction using proposed SPET motion-correction programs. The peak heart rate achieved during stress and the ratio of the heart rate immediately before SPET acquisition to the resting heart rate were determined to be independent predictors of patient motion during SPET acquisition. Cardiac motion changes were minimal at (13.3 +/- 2.2) min after cessation of exercise. The implications of these findings for the accuracy of SPET 201Tl require further investigation.