Regions of Interest–Based Discriminant Analysis of DaTSCAN SPECT and FDG-PET for the Classification of Dementia

Abstract

Neuroimaging is increasingly used to support the clinical diagnosis of patients with cognitive impairment. Dopamine transporter (DAT) imaging, such as DaTSCAN SPECT, tests the integrity of the nigrostriatal pathway, whereas FDG-PET identifies typical patterns of cortical and subcortical hypometabolism. The aim of this study was to assess the relative contribution of DAT and regional glucose metabolism imaging to the differential diagnosis. Twenty-seven subjects were investigated for neurodegenerative dementia associated with parkinsonism of variable severity by FDG-PET and DaTSCAN SPECT. They were grouped according to the clinically established diagnosis, including probable Alzheimer disease (5 subjects), corticobasal degeneration (6 subjects), Lewy body dementia (8 subjects), frontotemporal dementia (4 subjects), and Parkinson disease with dementia (4 subjects). Normalized FDG uptake and DAT uptake ratios were obtained by the BRASS software. We used a discriminant analysis with a stepwise method and a leave-one-out cross-validation. With the use of regional values of normalized FDG uptake, 85.2% and 55.6% of the patients were correctly classified by the discriminant analysis and the cross-validation, respectively. When DAT alone was considered, the results were 59.3% and 51.9%, whereas the combination of both DAT and normalized FDG uptake yielded 100% and 88.9% of accurate classifications. This automated analysis approach shows that the information provided by normalized FDG uptake and DAT is not redundant for the differential diagnosis of dementia and that taking into account both normalized FDG uptake and DAT uptake allows a better classification of individual patients. These results further support the usefulness of both modalities in the clinical workup of dementia.