Fig. 2

Visual summary of data collection and analysis. A Data collection and feature extraction. Multi-modal data is collected from the recruited patients: clinical-neuropsychological evaluations, genetic and biological data, EEG at rest and during memory and attention tasks (ERP). From the EEG signals, features are extracted by using several analyses, such as connectivity (top left), microstates (top right), spectral and ERP time course analyses. B The extracted features, candidate biomarkers of progression to AD, are submitted to the cross-validated machine learning framework. First, only informative features are selected (green squares), while the non-informative ones are discarded. The selected features are used as inputs to train a set of machine learning classifiers (e.g., an ANN is displayed) to determine whether: i) the subject is carrier of biological AD pathology; ii) will remain SCD or will progress towards MCI or AD dementia. Abbreviations: CSF = cerebrospinal fluid; EEG = electroencephalogram; ERP = event related potentials; 3 CVT = 3-choice vigilance task; SIR = standard image recognition task; SCD = subjective cognitive decline; MCI = mild cognitive impairment; AD = Alzheimer’s disease; BDNF = brain-derived neurotrophic factor; APOE = apolipoprotein E