Magnetic resonance imaging, computed tomographic, and positron emission tomographic studies of the brain provide complementary information, and many patients undergo more than one of these studies during the course of their diagnostic workup and treatment. A new technique for quantitative geometric correlation of such studies makes it possible to create integrated multimodality images by mapping features from one image onto an image obtained with another modality. The coordinate transformation between any pair of images is found by a semiautomatic algorithm for matching models of the patient's external surface as depicted in the two data sets. The resultant hybrid images, which combine complementary features of different studies, are often more useful for diagnosis and treatment planning than are the original single-modality images. The algorithm can also be used for spatial registration of baseline studies with follow-up images created with the same modality, which allows tracking of a lesion to detect subtle interval changes in size and shape. This technique can be applied to images acquired in routine clinical practice, since it is completely retrospective and does not necessitate special positioning or landmarking of the patient.