Fully automated structural MRI of the brain in clinical dementia workup.

Background The dementia syndrome has been regarded a clinical diagnosis but the focus on supplemental biomarkers is increasing. An automatic magnetic resonance imaging (MRI) volumetry method, NeuroQuant® (NQ), has been developed for use in clinical settings. Purpose To evaluate the clinical usefulness of NQ in distinguishing Alzheimer's disease dementia (AD) from non-dementia and non-AD dementia. Material and Methods NQ was performed in 275 patients diagnosed according to the criteria of ICD-10 for AD, vascular dementia and Parkinson's disease dementia (PDD); the Winblad criteria for mild cognitive impairment; the Lund-Manchester criteria for frontotemporal dementia; and the revised consensus criteria for Lewy body dementia (LBD). Receiver operating curve (ROC) analyses with calculation of area under the curve (AUC) and regression analyses were carried out. Results Forebrain parenchyma (AUC 0.82), hippocampus (AUC 0.80), and inferior lateral ventricles (AUC 0.78) yielded the highest AUCs for AD/non-dementia discrimination. Only hippocampus (AUC 0.62) and cerebellum (AUC 0.67) separated AD from non-AD dementia. Cerebellum separated AD from PDD-LBD (AUC 0.83). Separate multiple regression analyses adjusted for age and gender, showed that memory (CERAD 10-word delayed recall) (beta 0.502, P < 0.001) was more strongly associated to the hippocampus volume than the diagnostic distinction of AD versus non-dementia (beta -0.392, P < 0.001). Conclusion NQ measures could separate AD from non-dementia fairly well but generally poorer from non-AD dementia. Degree of memory impairment, age, and gender, but not diagnostic distinction, were associated to the hippocampus volume in adjusted analyses. Surprisingly, cerebellum was found relevant in separating AD from PDD-LBD.