Thedomestic pig is increasingly being used as an experimental model for brain imaging studies with positron emission tomography (PET). The recording of radiotracer uptake by PET gives functional and physiological information, but with poor spatial resolution. To date, anatomical regions of interest in pig brain have been defined in MR images obtained for each individual animal, because of the lack of a standard stereotaxic coordinate system for the pig brain. In order to define a stereotaxic coordinate system, we coregistered T1-weighted MR images from 22 male Göttingen minipigs and obtained a statistically defined surface rendering of the average minipig brain in which stereotaxic zero is defined by the position of the pineal gland. The average brain is now used as a target for registration of dynamic PET data, so that time-activity curves can be extracted from standard volumes of interest. In order to define these volumes, MR images from each individual pig were manually segmented into a total of 34 brain structures, including cortical regions, white matter, caudate and putamen, ventricular system, and cerebellum. The mean volumes of these structures had variances in the range of 10-20%. The 34 brain volumes were transformed into the common coordinate system and then used to generate surface renderings with probabilistic threshold greater than 50%. This probabilistic threshold gave nearly quantitative recovery of the mean volumes in native space. The probabilistic volumes in stereotaxic space are now being used to extract time-radioactivity curves from dynamic PET recordings.