Background: Neuroimaging has a reasonable accuracy to differentiate behavioral variant frontotemporal dementia (bvFTD) from other neurodegenerative disorders, its value for the differentiation of bvFTD among subjects with acquired behavioral disturbances is unknown. Objective: To determine the diagnostic accuracy of MRI, additional [18F]FDG-PET, and their combination for bvFTD among subjects with late onset behavioral changes. Methods: Patients with late onset behavioral changes referred to a memory clinic or psychiatric services were included. At baseline, 111 patients had a brain MRI scan and 74 patients received an additional [18F]FDG-PET when the MRI was inconclusive. The consensus diagnosis after two-year-follow-up was used as the gold standard to calculate sensitivity and specificity for baseline neuroimaging. Results: 27 patients had probable/definite bvFTD and 84 patients had a non-bvFTD diagnosis (primary psychiatric diagnosis or other neurological disorders). MRI had a sensitivity of 70% (95% CI 52–85%) with a specificity of 93% (95% CI 86–97%). Additional [18F]FDG-PET had a sensitivity of 90% (95% CI 66–100%) with a specificity of 68% (95% CI 56–79%). The sensitivity of combined neuroimaging was 96% (95% CI 85–100%) with a specificity of 73% (95% CI 63–81%). In 66% of the genetic FTD cases, MRI lacked typical frontotemporal atrophy. 40% of cases with a false positive [18F]FDG-PET scan had a primary psychiatric diagnosis. Conclusion: A good diagnostic accuracy was found for MRI and additional [18F]FDG-PET for bvFTD in patients with late onset behavioral changes. Caution with the interpretation of neuroimaging results should especially be taken in cases with a genetic background and in cases with a primary psychiatric differential diagnosis where [18F]FDG-PET is the only abnormal investigation.