Multimodal Fusion for Staging Classification in Alzheimer's Disease Using Extended parallel multilink joint ICA

Motivation: Combining multimodal MRI data may help to improve the accuracy of calcification in Alzheimer's disease(AD). Goal(s): By integrating structural MRI (sMRI), DTI, and functional MRI (fMRI) data, we aimed to discover biomarkers for AD calcification. Approach: Extended pml-jICA approach was applied for multimodal fusion (sMRI, DTI, and fMRI) to identify biomarkers for AD classification. Results: Loading parameters revealed significant differences among AD, mild cognitive impairment (MCI), subjective cognitive decline (SCD), and normal cognition (NC) in certain components. The accuracy of the four-class classification using SVM was 55%, and the accuracy for the binary classification of SCD and NC was 80%. Impact: The observed differences of loading parameters among AD, MCI, SCD, and NC suggested that the combination of multimodal data may provide useful biomarkers for improving the accuracy AD diagnosis.