The parameters of reconstruction iteration and subset of the ordered subset expectation maximization (OS-EM) method have a great impact on image quality. In a cerebral blood flow (CBF) examination, it is critical to produce consistent results when analyzing clinical data. We evaluated the number of iterations and subset of the OS-EM method to stabilize image quality using a normal case and an infarction case. In the maximum likelihood expectation maximization (ML-EM) method, it was confirmed that the convergence of an infarction case was somewhat delayed compared with a normal case. With the OS-EM method, we have obtained the same results. Based on the rCBF values, we postulate that the iteration value is more than 8 and the subset value is less than 10. Furthermore, multiplying the iteration by a subset value ranging from 50 to 60 helps in stabilizing the quality of CBF imaging.