Abstract

Diagnosis of Alzheimer’s disease (AD) at the early stage of the disease development is of greatclinical importance. Current clinical assessment that relies primarily on cognitive measures proveslow sensitivity and specificity. The fast growing neuroimaging techniques hold great promise.Research so far has focused on single neuroimaging modality. However, as different modalitiesprovide complementary measures for the same disease pathology, fusion of multi-modality datamay increase the statistical power in identification of disease-related brain regions. This isespecially true for early AD, at which stage the disease-related regions are most likely to be weak-effect regions that are difficult to be detected from a single modality alone. We propose a sparsecomposite linear discriminant analysis model (SCLDA) for identification of disease-related brainregions of early AD from multi-modality data. SCLDA uses a novel formulation that decomposeseach LDA parameter into a product of a common parameter shared by all the modalities and aparameter specific to each modality, which enables joint analysis of all the modalities andborrowing strength from one another. We prove that this formulation is equivalent to a penalizedlikelihood with non-convex regularization, which can be solved by the DC (difference of convexfunctions) programming. We show that in using the DC programming, the property of the non-convex regularization in terms of preserving weak-effect features can be nicely revealed. Weperform extensive simulations to show that SCLDA outperforms existing competing algorithms onfeature selection, especially on the ability for identifying weak-effect features. We apply SCLDAto the Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) images of49 AD patients and 67 normal controls (NC). Our study identifies disease-related brain regionsconsistent with findings in the AD literature.