Identifying pre-clinical Alzheimer's disease (AD) in subjects with mild cognitive impairment (MCI) is a major issue in clinical diagnosis. Establishing a combination of predictive markers from different fields of research might help in increasing the diagnostic accuracy. Aim of this study was to evaluate the potential role of 99mTc-ECD single photon emission computed tomography (SPECT) and memory scores in predicting conversion to AD in MCI subjects. Thirty-one MCI subjects underwent a clinical and neuropsychological examination, and a regional cerebral blood flow (rCBF) SPECT scan at baseline. Subjects had been followed periodically through 2 years in order to monitor the progression of cognitive symptoms. Canonical variate analysis of principal components was able to separate all subjects who converted to AD from those who remained stable, the former being characterized by a specific hypometabolic pattern, involving the parietal and temporal lobes, precuneus, and posterior cingulate cortex. Canonical correlation analysis of combined baseline memory deficits and rCBF SPECT images identified pre-clinical AD with a sensitivity and specificity of 77.8%. The pattern of hypoperfusion 99mTc-ECD SPECT and the severity of memory deficits predict the risk of progression to probable AD dementia in MCI subjects.