Using global constraints and dynamic programming, a new model-based segmentation algorithm was developed to determine myocardial borders and basal plane. The segmented image is transformed to a countrate polar map and the infarct size (I.S.) is determined by comparison with a reference polar map. In order to evaluate our method the algorithm was applied to heart phantoms, to software simulations and to animal studies. In the last experiments, Tc-99m Sestamibi was used as a perfusion agent. The total myocardial volume and infarct size of a Jasczack phantom were overestimated, especially when I.S. was expressed in absolute rather than relative values. It was proven by software simulations of cardiac Spect studies that those errors were mainly due to finite resolution effects causing a clear overestimation of myocardial thickness. Implementation of a constant thickness in the algorithm resulted in a much better correlation with actual values. In a dog experiment the size of total myocardial volume of the area at risk during occlusion and of the final infarct size after thrombolysis was correlated with the histologic values obtained by planimetry after TTC staining. In 8 studies, an excellent correlation between the histologic area at risk versus the estimated perfusion defects was obtained (r = 0.97). The automatic delineation of myocardial borders and valve plane was excellent even when perfusion defects were present. Manual intervention was only necessary in certain slices where a clear overlap between liver and myocardium was present in the dog studies. Segmental polar maps expressing count rate and volume information provided a visual and quantitative tool to evaluate the influence of thrombolysis in acute ligation experiments. It is concluded that the new algorithm is ready to be used in a clinical environment for the quantitative evaluation of perfusion defects after acute myocardial infarction and for the follow-up of the therapeutic strategy.