An adjustable computerized atlas of the human brain has been developed, which can be adapted to fit individual anatomy. It is primarily intended for positron emission tomography (PET) but may also be used for single photon emission CT, transmission CT, magnetic resonance imaging, and neuroimaging-based procedures, such as stereotactic surgery and radiotherapy. The atlas is based on anatomical information obtained from brains fixed in situ soon after death. All structures have been drawn in on digitized photos of slices from one cryosectioned brain. The definition and classification of the anatomical structures and divisions are in agreement with the standard textbooks of anatomy, and the nomenclature is that of the Nomina Anatomica of 1965. The boundaries of the cortical cytoarchitectonic areas (Brodmann areas) have been determined using information from several sources, since three-dimensional literature data on their distribution are incomplete, scarce, and partly contradictory. However, no analysis of the cytoarchitectonics of the atlas brain itself has been undertaken. At present the data base contains three-dimensional representations of the brain surface, the ventricular system, the cortical gyri and sulci, as well as the Brodmann cytoarchitectonic areas. The major basal ganglia, the brain stem nuclei, the lobuli of the vermis, and the cerebellar hemispheres are also included. The computerized atlas can be used to improve the quantification and evaluation of PET data in several ways. For instance, it can serve as a guide in selecting regions of interest. It may also facilitate comparisons of data from different individuals or groups of individuals, by applying the inverse atlas transformation to PET data volume, thus relating the PET information to the anatomy of the reference atlas rather than to the patient's anatomy. Reformatted PET data from individuals can thus be averaged, and averages from different categories or different functional states of patients can be compared.