The purpose of this work was to determine the extent to which automated quantitative volumetric MR neuroimaging can distinguish persons with behavioral variant frontotemporal dementia (bvFTD) from the more common Alzheimer’s diseases (AD).

The ability of volumetric brain MRI to be utilized clinically would be better realized if objective data could demonstrate its ability to distinguish common forms of dementia from one another. As bvFTD must often be distinguished from AD, the most common neurodegenerative dementia, applying this question to these disorders has a high potential to affect future clinical practice.

A total of 95 cognitive neurology patients were evaluated by a behavioral neurologist of which 16 met Consensus Criteria for bvFTD and 79 met established criteria for clinically probable AD. All participants received a sagittal T1 weighted volumetric acquisition on a 3T MR scanner and were analyzed by the Neuroreader software to compute brain volumes. Volumes for lobar structures and limbic areas were inputted into a discriminant analysis in SPSS with leave one out cross validation. This analysis produced an ROC and feature selection was also done separately.

Clinically available, quantitative volumetric MR neuroimaging combined with a predictive analytic approach can distinguish bvFTD from AD with a high level of accuracy, sensitivity and specificity. Future work will evaluate such features in distinguishing other possible etiologies of dementia.