Automatic Classification of Alzheimers Disease vs. Frontotemporal Dementia: A Decision Tree Approach with FDG-PET

We introduce a novel approach for the automatic classification of FDG-PET scans of subjects with Alzheimers Disease (AD) and Frontotemporal dementia (FTD). Unlike previous work in the literature which focuses on principal component analysis and predefined regions of interest, we propose the use of decision tree learning combined with empirically determined regions of interest as attributes. The advantages of this approach are two-fold. First, empirically determining regions of interest for distinguishing between these two diseases is relevant for clinical medical practice. Second, we illustrate that the proposed method provides better classification accuracy compared to other methods on a group of 48 autopsy confirmed AD and FTD patients. Automatic Classification of Alzheimer’s Disease vs. Frontotemporal Dementia: A Decision Tree Approach with FDG-PET Neda Sadeghi, Tolga Tasdizen, Norman L. Foster, Angela Y. Wang, Satoshi Minoshima, Andrew P. Lieberman 1 School of Computing, University of Utah, Salt Lake City, UT 84112 2 Center for Alzheimer’s Care, Imaging and Research, University of Utah, Salt Lake city, UT 84112 3 School of Medicine, University of Washington, Seattle, WA 98195 4 Department of Pathology, University of Michigan, Ann Arbor, MI 48109 Abstract. We introduce a novel approach for the automatic classification of FDG-PET scans of subjects with Alzheimer’s Disease (AD) and Frontotemporal dementia (FTD). Unlike previous work in the literature which focuses on principal component analysis and predefined regions of interest, we propose the use of decision tree learning combined with empirically determined regions of interest as attributes. The advantages of this approach are two-fold. First, empirically determining regions of interest for distinguishing between these two diseases is relevant for clinical medical practice. Second, we illustrate that the proposed method provides better classification accuracy compared to other methods on a group of 48 autopsy confirmed AD and FTD patients. We introduce a novel approach for the automatic classification of FDG-PET scans of subjects with Alzheimer’s Disease (AD) and Frontotemporal dementia (FTD). Unlike previous work in the literature which focuses on principal component analysis and predefined regions of interest, we propose the use of decision tree learning combined with empirically determined regions of interest as attributes. The advantages of this approach are two-fold. First, empirically determining regions of interest for distinguishing between these two diseases is relevant for clinical medical practice. Second, we illustrate that the proposed method provides better classification accuracy compared to other methods on a group of 48 autopsy confirmed AD and FTD patients.