Dementia is an important cause of morbidity and mortality worldwide and encompasses a very heterogenous group of disease processes. Positron emission tomography (PET) of the brain using fluorodeoxyglucose (FDG) is a useful modality for differentiating types of dementia. Because FDG does not bind to pathologic proteins, FDG-PET requires that the reader recognize characteristic patterns of glucose hypometabolism to identify pathology. These patterns have been documented in the literature for both primary neurodegenerative disorders and secondary causes of dementia. This article presents an algorithm for organizing these findings and systematically applying them to interpret FDG-PET brain imaging for dementia.