Figure 1. Schematic diagram of the proposed classification, independent validation, and prediction framework. First, we calculated each individual CARE index score for MCI subjects in the ADNI and NADS datasets. Second, we utilized a CARE index score (A score on the CARE index is equivalent to the subject’s disease stage.) to classify N-MCI and P-MCI subjects in the ADNI dataset. Third, we applied the ‘CARE index stage “classifier” determined from the ADNI dataset to predict the conversion of MCI subjects in the NADS dataset. The ROC curve was used to assess the performance of the CARE index stage classifier and the CARE index stage prediction classifier, respectively. In addition, we assessed the performance differences of CARE index stage classification and prediction and original indices (AVLT, MMSE, GM, and FC indices) by comparing these ROCs across datasets. Abbreviations: MCI, mild cognitive impairment; ADNI, Alzheimer’s Disease Neuroimaging Initiative; NADS, Nanjing Aging and Dementia Study; AD, Alzheimer’s disease; MMSE, Mini-Mental State Examination; AVLT, Rey Auditory Verbal Learning Test; MRI, magnetic resonance imaging; AAL, automated anatomical labeling; GM, grey matter; BOLD, blood oxygenation level dependent; FC, functional connectivity; CARE, characterizing AD risk event; ROC, receiver operating characteristic; P-MCI, progressive MCI, including MCI subjects who progressed to AD-type dementia at the three-year follow-up; N-MCI, no-progressive MCI, including MCI subjects who had not progressed to dementia at the three-year follow-up.