AUTHOR=Khatri Uttam , Kwon Goo-Rak 

TITLE=Alzheimer’s Disease Diagnosis and Biomarker Analysis Using Resting-State Functional MRI Functional Brain Network With Multi-Measures Features and Hippocampal Subfield and Amygdala Volume of Structural MRI

JOURNAL=Frontiers in Aging Neuroscience

VOLUME=Volume 14 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/aging-neuroscience/articles/10.3389/fnagi.2022.818871

DOI=10.3389/fnagi.2022.818871

ISSN=1663-4365

ABSTRACT=Accurate diagnosis of the early stage of Alzheimer’s disease (AD) is essential but very challenging. The objective of this study was to properly utilize the biomarkers for diagnosis of AD based on merging resting-state functional MRI brain networks and voxel features. The fMRI time series data have a potentiality of the specific numerical information as well as dynamic temporal information. Voxel and graphical analysis have gained the popularities to analyzing neurodegenerative disorders such as mild cognitive impairment (MCI) and AD. So far, these methods have been utilized separately for classification of AD and its prodromal stage MCI. In the major of studies classification of mild cognitive impairment which do not convertible for certain period known as stable MCI (MCIs) and which are convertible to AD (MCIc) is less commonly reported and have inconsistence results. In this study, we tested the efficacy of a purposed classification framework to identified AD and the stable mild cognitive impairment (MCIs) from convertible by utilizing the efficient features derived from functional brain networks of frequency band (0.01-0.027) at resting state and the voxel-based features ALFO, FFLO and ReHo. Pearson correlation coefficient to measuring functional brain network has been proof to be an efficient means to diagnose AD. Graphical theory was performed to calculate nodal features [nodal degree (ND), nodal path length (NL), and between centrality (BC)] as a graphical feature and analyze the relationship between changes in network connectivity. Subsequently, we extracted three-dimensional patterns calculating regional coherence, we performed univariate statistical t-test to generate 3-D mask that retained voxels showing significant changes. Finally, we implemented and compare the different features selection algorithm to integrate the brain networks and voxel features into one form to optimize the classifier, we used SVM classifiers. Experimental results verify the effectiveness of our proposed method; combination of brain network with multiple measure of rs-fMRI significantly enhance in accuracy over other approaches on AD. The accuracy obtained by the proposed method were reported on binary classification. More importantly classification results of less commonly reported group MCIs vs MCIc improved significantly.