• connectivity
• event related potentials
• machine learning
• mild cognitive impairment
• subjective cognitive decline

This study investigated neural dynamics from Event Related Potentials (ERPs) recorded from patients with Subjective Cognitive Decline (SCD) and Mild Cognitive Impairment (MCI) during a 3-Choice Vigilance Task (3-CVT). Both SCD and MCI are part of the Alzheimer’s disease (AD) spectrum, standing as possible preliminar stages of the disease. While clinical features of prodromal cognitive decline conditions serve as diagnostic criteria, they do not provide any insights about the risk of progression in the spectrum. The analysis of brain neural signals could fill this gap, by highlighting potential biomarkers associated with a higher risk of disease evolution. In this framework, the present thesis is intended to show (i) if patients affected with SCD and MCI exhibit connectivity differences and (ii) if there are within-group connectivity differences that could drive the prediction of the disease progression. As a result, functional, effective and graph theory connectivity metrics unveiled significant differences between the two clinical conditions, serving as features for machine learning-based classification. Linear Discriminant Analysis reached an 80% balanced accuracy and the patients who were mismatched by the connectivity-based classification showed ERP’s components features closely resembling those of the other class. These findings underscored network features in identifying cognitive decline conditions and the potential role of connectivity in predicting a clinical worsening for a more appropriate patient’s care, providing a valuable context within the broader research protocol of the PREVIEW project.