Supervised Classification of Resting-State fMRI Biomarkers for Early Detection of Alzheimer's Disease

Alzheimer's disease (AD) is an irreversible neurodegenerative disorder. Resting-state functional magnetic resonance imaging (rs-fMRI) provides several metrics that capture neural activity, including the amplitude of low-frequency fluctuations (ALFF), fractional ALFF (fALFF), regional homogeneity (ReHo), and voxel-mirrored homotopic connectivity. However, the integration of these metrics with graph-theoretical metrics and machine learning techniques for accurate early detection of AD is still not well established. To address this problem, this study systematically combined the above analysis methods and employed multiple machine learning models to stage the different phases of AD. The results showed that, from the cognitively normal stage to the early mild cognitive impairment stage, ALFF, fALFF, and ReHo were altered, classification with logistic regression achieved an area under the curve (AUC) of 9 7. 5%. From the late mild cognitive impairment stage to AD, significant changes were observed in smallworldness metrics, and logistic regression reached an AUC of 96.8%. These findings deepen the understanding of AD's neural mechanisms and highlight functional metrics based on rs-fMRI as promising biomarkers for early detection.