Group sparse dictionary learning and inference for resting-state fMRI analysis of Alzheimer's disease

Abstract

Recent research in neuroscience fields has revealed that the complicated experiments and the resting state responses are largely focused on by many researchers. In such tasks, however, we usually can not establish accurate hypothesis or even there is no experiment paradigm in the resting state measurement. And, it has been a core challenge to develop a universal method for squeezing out the meaningful information in the brain without any constraints of prior information or hypotheses on the functional magnetic resonance imaging (fMRI) data we measured. In contrast to the general linear model (GLM) that requires the hypothesis of expected response to certain tasks which is widely known and used, data-driven analysis methods have attracted attention thanks to their nature capable of blind separation of sources. The representative methods of these are principal component analysis (PCA) and independent component analysis (ICA). Although these classical data-driven analysis methods, PCA and ICA, have long been attractive methods in terms of providing analysis tools with no prior information about tasks of the fMRI data, which both have constraints that sources are assumed to be orthogonal and independent, respectively. These constrains often lead the restriction on explaining the underlying activities of brain in fMRI data. Recently, there has been increased interest in the use of neuroimaging techniques to investigate what happens in a brain at rest. Functional imaging studies have revealed that the default-mode network activity is disrupted in diseases such as Alzheimer`s (AD). However, there is no consensus, as yet, on the choice of analysis method for the application of resting-state analysis. A novel group analysis tool for data-driven resting state fMRI analysis using group sparse dictionary learning and mixed model is presented along with the promising indications of Alzheimer`s disease progression. Instead of using independency assumption as in popular I...