In a prospective study of more than 200 cases of dementia and 119 controls, annual technetium-99m–hexamethyl-propylene amineoxime (99mTc-HMPAO) single-photon emission computed tomography (SPECT) and annual medial temporal lobe (MTL) oriented X-ray computed tomography (CT) have been used to evaluate the diagnostic potential of functional and structural neuroimaging in the differential diagnosis of dementia. Some subjects have had up to 7 annual evaluations. So far, of 151 who have died, 143 (95%) have come to necropsy. Histology is known for 118, of whom 80 had Alzheimer's disease (AD), 24 had other “non-AD” dementias, and 14 controls with no cognitive deficit in life also had no significant central nervous system pathology. To compare the findings in the dementias with the profile of structural and functional imaging in the cognitively normal elderly, scan data from 105 living, elderly controls without cognitive deficit have also been included in the analysis. All clinical diagnoses were according to National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) and the Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; DSM-III-R) criteria, and all histopathological diagnoses according to the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) criteria. Early data from this cohort have suggested that the combination of both MTL atrophy seen on CT with parietotemporal hypoperfusion on SPECT may predict the pathology of AD. The diagnostic sensitivity, specificity, accuracy, and positive and negative predictive values of the NINCDSADRDA and DSM-III-R criteria could be assessed in this cohort against the gold standard of histopathology. The diagnostic potential of CT evidence of MTL atrophy alone, SPECT evidence of parietotemporal hypoperfusion alone, and the combination of both of these scan changes in the same individual could then be compared against the diagnostic accuracy of clinical operational criteria in the pathologically confirmed cases. Furthermore, all of these modalities could be compared with the diagnostic accuracy of apolipoprotein E4 (Apo E4) genotyping to predict AD in the histopathologically confirmed cohort. In this population, NINCDS “probable-AD” was 100% specific, 49% sensitive, and 66% accurate; “possible-AD” was only 61% specific, but 93% sensitive and 77% accurate; and the combination of both “probable-AD” and “possible-AD” was 61% specific, 96% sensitive, and 85% accurate. DSM-111-R criteria were 51% sensitive, 97% specific, and 66% accurate. In the same cases and including the 105 living, elderly controls, the diagnostic accuracy of the Oxford Project to Investigate Memory and Aging (OPTIMA) scanning criteria showed CT alone to be 85% sensitive, 78% specific, and 80% accurate; SPECT alone had 89% sensitivity, 80% specificity, and 83% accuracy; and the combination of the two was 80% sensitive, 93% specific, and 88% accurate. The Apo E4 genotype was 74% sensitive but yielded 40% false positives in the histologically confirmed series. The diagnostic accuracy afforded by this method of CT and SPECT used alone is better than that of any established clinical criteria and reveals that the combination of MTL atrophy and parietotemporal hypoperfusion is common in AD, much less common in other dementias, and rare in normal controls. In the NINCDS-ADRDA criteria “possible-AD” cases, the combination of CT and SPECT findings alone were better in all diagnostic indices than the presence of Apo E4 alone in predicting AD. The frequent occurrence of MTL atrophy in AD and also in other “non-AD” dementias later in the course of the disease suggests the concept of medial temporal lobe dementia. This could explain some of the overlap of clinical profiles in the dementias, particularly as the dementia progresses, making clinical differential diagnosis difficult. In this context, the use of SPECT can significantly enhance specificity.