RT info:eu-repo/semantics/article
T1 Integrative and interpretable framework to unveil the neurophysiological fingerprint of Alzheimer’s disease and mild cognitive impairment: A machine learning-SHAP approach
A1 Gutierrez De Pablo, Victor
A1 Herrero Tudela, María
A1 Sandonís Fernández, Marina
A1 Poza Crespo, Jesús
A1 Maturana Candelas, Aaron
A1 Rodríguez González, Víctor
A1 Tola Arribas, Miguel Ángel
A1 Cano, Mónica
A1 Hoshi, Hideyuki
A1 Shigihara, Yoshihito
A1 Hornero Sánchez, Roberto
A1 Gómez, Carlos
K1 Alzheimer’s disease
K1 Mild cognitive impairment
K1 Machine learning
K1 SHAP
K1 Magnetoencephalography
K1 Electroencephalography
K1 Neurophysiological fingerprint
K1 32 Ciencias Médicas
K1 33 Ciencias Tecnológicas
AB Dementia and mild cognitive impairment (MCI) due to Alzheimer’s disease (AD) are neurological pathologiesassociated with disruptions in brain electromagnetic activity, typically studied using magnetoencephalography(MEG) and electroencephalography (EEG). To quantify diverse brain properties, different families of param-eters can be computed from MEG and EEG (i.e., spectral, non-linear, morphological, functional connectivity,or network structure and organisation). However, studying these characteristics separately overlooks the com-plex nature of brain activity. Integrative frameworks can be useful to unveil the intricate neurophysiologicalfingerprint, as well as to characterise pathological conditions comprehensively. To that purpose, data fusionmethodologies are crucial, despite their interpretational challenges. In this study, Machine Learning (ML) mod-els were trained to discriminate between groups of severity, whereas the SHapley Additive eXplanations (SHAP)algorithm was afterwards utilised to assess the relevance of the input characteristics into the output classifica-tion. Three databases were analysed: MEG (55 healthy controls, HC, 42 MCI patients, and 86 AD patients), EEG1(51 HC, 52 MCI, and 100 AD), and EEG2 (45 HC, 69 MCI, and 82 AD). The best results for the three-class classi-fication problem were obtained by Gradient Boosting for the MEG database: 3-class Cohen’s kappa coefficient of0.5452 and accuracy of 72.63 %. Afterwards, using SHAP on Gradient Boosting, it has been shown that spectralfeatures were identified as highly relevant across all databases. Furthermore, morphology measures presentedhigh relevance for the MEG database, whereas EEG1 and EEG2 databases showed functional connectivity andmultiplex organisation measures, respectively, as relevant subgroups of parameters. Finally, commonly relevantfeatures across databases were selected using SHAP to generate the neurophysiological fingerprints of AD andMCI. This study highlights the relevance of different MEG and EEG parameters in characterising neurologicalpathologies. The proposed framework, based on MEG and EEG, can be used to generate interpretable, robust,and accurate neurophysiological fingerprints of AD and MCI.
PB Elsevier
SN 0208-5216
YR 2025
FD 2025
LK https://uvadoc.uva.es/handle/10324/78210
UL https://uvadoc.uva.es/handle/10324/78210
LA eng
NO Biocybernetics and Biomedical Engineering, 2025, vol. 45, n. 3, p. 438-450
NO Producción Científica
DS UVaDOC
RD 31-oct-2025