Medial temporal lobe atrophy determined by temporal lobe oriented computed tomography (CT), 1 year before death, is strongly associated with histopathologically confirmed Alzheimer's disease (AD). The aim of this study was to assess the diagnostic accuracy of medial temporal lobe measurement for the diagnosis of AD in patients referred to a memory disorders clinic, especially those at an early stage of the disease. CT oriented to the temporal lobe was performed in 333 subjects aged 41-93 years consecutively recruited in a Memory Disorders Clinic: 124 had probable AD, Mini Mental State score (MMS) = 17 (8); 50 possible AD [MMS = 21 (5)]; and 119 patients had miscellaneous memory disorders [MMS = 22 (7): frontotemporal lobe dementia, subcortical dementia, cortical Lewy body disease, vascular dementia, Korsakoff syndrome, focal atrophy, etc.]. There were also 19 anxious/depressed patients [MMS = 29 (1)] with normal performance on memory tests, and 21 controls. The minimum width of the medial temporal lobe was measured. The best cut-off to distinguish AD patients from non-AD patients was 11.5 mm, in agreement with data in the literature. At this threshold, 84% of probable AD patients had a positive test and 90% of controls and anxious/depressed patients had a negative test. For the diagnosis of probable AD, sensitivity of the measurement was 0.81, specificity 0.95, predictive positive value 0.99, predictive negative value 0.45, and diagnostic accuracy 0.83. The test was positive in half the possible AD patients, and half those with miscellaneous memory disorders. It was negative in all anxious/depressed patients. Therefore, temporal lobe oriented CT might be a valuable tool for assessment of medial temporal lobe atrophy in AD routine practice.