To investigate a potential application of ordered subsets expectation maximization (OSEM) algorithm for clinical [15O]H2O PET studies, region of interest (ROI) measurements were performed on both images with OSEM and filtered back projection (FBP). Forty OSEM images were reconstructed with variable combinations of numbers of the subset (1-40) and iteration times (2-12). PET scans were acquired using a PET/CT scanner (Discovery ST Elite, GE), and 3T-MRI images were obtained for fusion images. The mean values were measured on the frontal cortical regions in the middle cerebral artery distribution. Differences of the values between the OSEM and FBP were evaluated as %Error. Relationship between ROI mean values and the iteration times was investigated on the OSEM images. The smallest %Error 0.4% was measured in the combination of the subset number 10 and iteration times 8 [10, 8], and in that of [28, 2].The mean values were stable with iteration number 8 or more. OSEM image with [28, 2] was reconstructed in a shorter time (2.5 min) than that with [10, 8] (6 min). OSEM image with [28, 2] was superior to that with [10, 8] in the qualitative evaluation. The mean values on OSEM images with [28, 2] were comparable with those on FBP images with little artifacts and higher spatial resolution. OSEM with optimal parameter setting seemed applicable for both quantitative and qualitative [15O]H2O PET studies.