A comparison of classification methods for differentiating fronto-temporal dementia from Alzheimer's disease using FDG-PET imaging

摘要: Flurodeoxyglucose positron emission tomography (FDG-PET) is being explored to determine its ability differentiate between a diagnosis of Alzheimer's disease (AD) and fronto-temporal dementia (FTD). We have examined statistical discrimination procedures help achieve this purpose compared the results visual ratings FDG-PET images. The methods are applied data set 48 subjects with autopsy confirmed diagnoses AD or FTD (these come from multi-centre collaborative study funded by National Coordinating Center). images composed thousands voxels (volume elements) so one left situation where there vastly more variables than subjects. Therefore, it necessary perform reduction before procedure can be applied. Approaches using both entire image summary statistics calculated on number volumes interest (VOI) examined. performed techniques principal components analysis (PCA) partial least-squares (PLS) then used linear discriminant (LDA), quadratic (QDA) logistic regression (LR) classify as having FTD. Some these diagnostic accuracy (as assessed leave-one-out cross-validation) that similar expert raters. Methods PLS appear successful. Averaging VOI may also helpful.