We developed cerebral perfusion Z score map (Z map) images using H(2)15O and positron emission tomography (PET), and examined their use in diagnosing Alzheimer's disease (AD). Cerebral blood flow (CBF) images were obtained using the PET and H(2)15O autoradiographic method. The best region for normalising the CBF value to remove individual variantion was determined. Then CBF images were transformed to Talairach's standard space, and each pixel value of an individual's image set was normalized to the mean value of the sensorimotor area. Based on the CBF images of 20 normal volunteers, normative mean and standard deviation (SD) CBF images were constructed. Then, each pixel value of the axial CBF images in 28 patients with probable AD and 10 further normal volunteers was converted to a Z score (Z = [normal mean--individual value]/normal SD). A Z map, showing pixels exceeding a threshold of Z score > 2 on MRI of standardised anatomical space was demonstrated. These 38 Z maps were interpreted by four radiologists. When regions in the temporoparietal area were found with Z scores > 2, the subject was diagnosed as having AD. A receiver operating characteristic (ROC) analysis was performed to compare the conventional CBF images and Z maps. The diagnostic performance of the Z map was superior to that of visual inspection of conventional CBF images (mean areas under the ROC curve of the four radiologists were 0.946 and 0.584, respectively). These results indicate that a Z map obtained in this way is superior to conventional PET for diagnosing AD.