Joint ICA Based Auxiliary Diagnostics Of Alzheimer’s Disease

Alzheimer’s disease (AD) is a very common neurodegenerative disease in elderpeople. In China, as the growth of aging population accelerating, more and morepeople get AD. It also brings difficulties to economy. In these days, the MagneticResonance Imaging (MRI) and Positron Emission Tomography (PET) technologies arebecoming more and more mature. The comparison between anatomical brain changesand physiological changes like metabolism is the key to diagnose AD.Independent Component Analysis (ICA) is a new method for data processing. ICAis able to access mixing linear independent component from hidden signals. Recently itis widely used in medical image processing. No matter for MRI or PET images, ICA isa popular method for several imaging modalities. For one definite modality, ICAmethod can be used to decompose different independent components. Each componentcan be assessed alone. However, in this research, more than one imaging modalitiesneed to be co-assessed and only ICA method cannot achieve that. So joint ICA is used,instead of ICA, to jointly analyze data. Grey matter image is a kind of cerebralstructural image and can be used to assess grey matter atrophy. FDG-PET can beassessed about metabolism. Joint ICA could help to get association between these twokinds of images as well as each modality’s information.The raw images are acquired from ADNI and preprocessed based on SPM8Toolbox of Matlab. Moreover, preprocessed images also need to transform to changingrate image for joint ICA processing. Finally, the images of independent componentsmap to standard brain template. The outcome implies: For HC and MCI subjects, jointpatterns of covariation between the rates of decline in GM volume and FDG-PETmetabolism are different. Compared to HC, changing rate become faster and affectedbrain region become larger for MCI. Besides, the affected brain region is also differentbetween two stages. Compared Aβ+to Aβ-group, the changes in brain is alsodifferent.The significance and innovation of this research are:①Joint patterns ofcovariation between two modalities can be acquired by joint ICA method. Compared totraditional ICA method, joint analysis with correlated information brings moreevidence to diagnose the risk of AD and also improve the accuracy of diagnosis.②Due to few related longitudinal research, the longitudinal research in our research, especially the joint longitudinal research combined two modalities, is crucial to knowthe changes and rules in early AD.