Deep Learning Based on Electroencephalogram Biomarkers for Identifying Early Preclinical Stages of Alzheimer's Disease

• Grant number: 25/25034-2
• Support Opportunities: Scholarships abroad - Research
• Start date: September 01, 2026
• End date: August 31, 2027
• Field of knowledge: Engineering - Biomedical Engineering - Bioengineering
• Principal Investigator: Francisco José Fraga da Silva
• Grantee: Francisco José Fraga da Silva
• Host Investigator: Claudio Babiloni
• Host Institution: Centro de Engenharia, Modelagem e Ciências Sociais Aplicadas (CECS). Universidade Federal do ABC (UFABC). Santo André , SP, Brazil
• Institution abroad: Università degli Studi di Roma La Sapienza, Italy

Abstract

This FAPESP Overseas Research Grant project proposes the development of artificial intelligence methods based on Deep Learning (DL) to assess the early pre-clinical stages of Alzheimer's Disease (AD) from electroencephalogram (EEG) signals. The focus is to distinguish between cognitively preserved healthy older adults, older adults with subjective memory complaints (SCM) and older adults with mild cognitive impairment (MCI).The research is based on consolidated EEG databases from the INSIGHT-preAD study and PDWAVES Consortium (www.pdwaves.eu), provided by the Sapienza University of Rome. The project explores the relationship between cognitive reserve, brain rhythms, and functional connectivity, searching for neurophysiological patterns indicative of vigilance disturbances in those patients. Advanced EEG preprocessing techniques will be applied, including automated artifact removal and feature extraction and selection (such as power spectral density and brain source connectivity). Subject classification will be performed using Transformers deep neural networks, a recent and promising approach in temporal signal processing. The Leave-One-Subject-Out (LOSO) cross-validation paradigm will ensure greater robustness and generalizability of the model evaluation outcomes. The project aims to contribute to the advancement of non-invasive, accessible and low-cost biomarkers to support the instrumental assessment of vigilance disturbances of the early stages of AD, complementing expensive or invasive methods currently used (magnetic resonance imaging, positron emission tomography, cerebrospinal fluid extraction). We expect to obtain classifiers with high accuracy, sensitivity, and specificity, capable of supporting clinical decisions. In addition to its scientific impact, the project promotes international cooperation and knowledge transfer between the Federal University of ABC (UFABC) and the Sapienza University of Rome, consolidating the use of DL in neuroscience applied to cognitive aging. (AU)