PURPOSE: Injected doses are difficult to optimize for exercise SPECT since they depend on the myocardial fraction of injected activity (MFI) that is detected by the camera. The aim of this study was to analyse the factors affecting MFI determined using a cardiac CZT camera as compared with those determined using conventional Anger cameras. METHODS: Factors affecting MFI were determined and compared in patients who had consecutive exercise SPECT acquisitions with (201)Tl (84 patients) or (99m)Tc-sestamibi (87 patients) with an Anger or a CZT camera. A predictive model was validated in a group of patients routinely referred for (201)Tl (78 patients) or (99m)Tc-sestamibi (80 patients) exercise CZT SPECT. RESULTS: The predictive model involved: (1) camera type, adjusted mean MFI being ninefold higher for CZT than for Anger SPECT, (2) tracer type, adjusted mean MFI being twofold higher for (201)Tl than for (99m)Tc-sestamibi, and (3) logarithm of body weight. The CZT SPECT model led to a +1 ± 26% error in the prediction of the actual MFI from the validation group. The mean MFI values estimated for CZT SPECT were more than twofold higher in patients with a body weight of 60 kg than in patients with a body weight of 120 kg (15.9 and 6.8 ppm for (99m)Tc-sestamibi and 30.5 and 13.1ppm for (201)Tl, respectively), and for a 14-min acquisition of up to one million myocardial counts, the corresponding injected activities were only 80 and 186 MBq for (99m)Tc-sestamibi and 39 and 91 MBq for (201)Tl, respectively. CONCLUSION: Myocardial activities acquired during exercise CZT SPECT are strongly influenced by body weight and tracer type, and are dramatically higher than those obtained using an Anger camera, allowing very low-dose protocols to be planned, especially for (99m)Tc-sestamibi and in non-obese subjects.