AUTHOR=Zhu Qixiao , Wang Yonghui , Zhuo Chuanjun , Xu Qunxing , Yao Yuan , Liu Zhuyun , Li Yi , Sun Zhao , Wang Jian , Lv Ming , Wu Qiang , Wang Dawei TITLE=Classification of Alzheimer’s Disease Based on Abnormal Hippocampal Functional Connectivity and Machine Learning JOURNAL=Frontiers in Aging Neuroscience VOLUME=Volume 14 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/aging-neuroscience/articles/10.3389/fnagi.2022.754334 DOI=10.3389/fnagi.2022.754334 ISSN=1663-4365 ABSTRACT=Objectives: Alzheimer’s disease (AD) is one type of neurodegenerative disease characterized by progressive memory impairment and cognitive ability decline. Mild cognitive impairment (MCI) has been implicated as a preclinical phase of AD. Although studies have showed abnormal functional connectivity (FC) in AD and MCI, to distinguish AD, MCI and normal aging controls (NC) between each other, especially the latter two is still challenging. Methods: We hypothesized that the functional connectivity from hippocampus in AD and MCI is abnormal, and this abnormal FCs could play a significance role in classifying the different stages of AD. Five-dimension reduction/classification methods were employed using hippocampus-derived FC strength as input features from totally 127 participants, the classification performance was compared between any two groups of AD, MCI and NC. Results: We found that the hippocampus of AD and MCI have significant abnormal connectivity with left insula, left thalamus, cerebellum, right lingual gyrus, posterior cingulate cortex (PCC) and precuneus. SVM coordinated with Sparse Principal Component Analysis (SPCA) achieved the best performance, obtaining an classification accuracy of 82.02% (AD vs. NC), 81.33% (MCI vs. NC), 81.08% (AD vs. MCI). Conclusions: Results suggest that hippocampus-seed-based FCs do demonstrate significant group differences among AD, MCI and NC, which when combined with popular machine learning methods can also improve AD differential diagnosis remarkably, especially between MCI and NC.