TY - JOUR
T1 - Classification of healthy subjects and Alzheimer's disease patients with dementia from cortical sources of resting state EEG rhythms
T2 - A study using artificial neural networks
AU - Triggiani, Antonio Ivano
AU - Bevilacqua, Vitoantonio
AU - Brunetti, Antonio
AU - Lizio, Roberta
AU - Tattoli, Giacomo
AU - Cassano, Fabio
AU - Soricelli, Andrea
AU - Ferri, Raffaele
AU - Nobili, Flavio
AU - Gesualdo, Loreto
AU - Barulli, Maria Rosaria
AU - Tortelli, Rosanna
AU - Cardinali, Valentina
AU - Giannini, Antonio
AU - Spagnolo, Pantaleo
AU - Armenise, Silvia
AU - Stocchi, Fabrizio
AU - Buenza, Grazia
AU - Scianatico, Gaetano
AU - Logroscino, Giancarlo
AU - Lacidogna, Giordano
AU - Orzi, Francesco
AU - Buttinelli, Carla
AU - Giubilei, Franco
AU - Del Percio, Claudio
AU - Frisoni, Giovanni B.
AU - Babiloni, Claudio
PY - 2017
Y1 - 2017
N2 - Previous evidence showed a 75.5% best accuracy in the classification of 120 Alzheimer's disease (AD) patients with dementia and 100 matched normal elderly (Nold) subjects based on cortical source current density and linear lagged connectivity estimated by eLORETA freeware from resting state eyes-closed electroencephalographic (rsEEG) rhythms (Babiloni et al., 2016a). Specifically, that accuracy was reached using the ratio between occipital delta and alpha1 current density for a linear univariate classifier (receiver operating characteristic curves). Here we tested an innovative approach based on an artificial neural network (ANN) classifier from the same database of rsEEG markers. Frequency bands of interest were delta (2-4 Hz), theta (4-8 Hz Hz), alpha1 (8-10.5 Hz), and alpha2 (10.5-13 Hz). ANN classification showed an accuracy of 77% using the most 4 discriminative rsEEG markers of source current density (parietal theta/alpha 1, temporal theta/alpha 1, occipital theta/alpha 1, and occipital delta/alpha 1). It also showed an accuracy of 72% using the most 4 discriminative rsEEG markers of source lagged linear connectivity (inter-hemispherical occipital delta/alpha 2, intra-hemispherical right parietal-limbic alpha 1, intra-hemispherical left occipital-temporal theta/alpha 1, intra-hemispherical right occipital-temporal theta/alpha 1). With these 8 markers combined, an accuracy of at least 76% was reached. Interestingly, this accuracy based on 8 (linear) rsEEG markers as inputs to ANN was similar to that obtained with a single rsEEG marker (Babiloni et al., 2016a), thus unveiling their information redundancy for classification purposes. In future AD studies, inputs to ANNs should include other classes of independent linear (i.e., directed transfer function) and non-linear (i.e., entropy) rsEEG markers to improve the classification.
AB - Previous evidence showed a 75.5% best accuracy in the classification of 120 Alzheimer's disease (AD) patients with dementia and 100 matched normal elderly (Nold) subjects based on cortical source current density and linear lagged connectivity estimated by eLORETA freeware from resting state eyes-closed electroencephalographic (rsEEG) rhythms (Babiloni et al., 2016a). Specifically, that accuracy was reached using the ratio between occipital delta and alpha1 current density for a linear univariate classifier (receiver operating characteristic curves). Here we tested an innovative approach based on an artificial neural network (ANN) classifier from the same database of rsEEG markers. Frequency bands of interest were delta (2-4 Hz), theta (4-8 Hz Hz), alpha1 (8-10.5 Hz), and alpha2 (10.5-13 Hz). ANN classification showed an accuracy of 77% using the most 4 discriminative rsEEG markers of source current density (parietal theta/alpha 1, temporal theta/alpha 1, occipital theta/alpha 1, and occipital delta/alpha 1). It also showed an accuracy of 72% using the most 4 discriminative rsEEG markers of source lagged linear connectivity (inter-hemispherical occipital delta/alpha 2, intra-hemispherical right parietal-limbic alpha 1, intra-hemispherical left occipital-temporal theta/alpha 1, intra-hemispherical right occipital-temporal theta/alpha 1). With these 8 markers combined, an accuracy of at least 76% was reached. Interestingly, this accuracy based on 8 (linear) rsEEG markers as inputs to ANN was similar to that obtained with a single rsEEG marker (Babiloni et al., 2016a), thus unveiling their information redundancy for classification purposes. In future AD studies, inputs to ANNs should include other classes of independent linear (i.e., directed transfer function) and non-linear (i.e., entropy) rsEEG markers to improve the classification.
KW - Alzheimer's disease (AD)
KW - Artificial neural networks (ANNs)
KW - Electroencephalography (EEG)
KW - Exact low-resolution brain electromagnetic tomography (eLORETA)
KW - Linear lagged connectivity
UR - http://www.scopus.com/inward/record.url?scp=85011891104&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85011891104&partnerID=8YFLogxK
U2 - 10.3389/fnins.2016.00604
DO - 10.3389/fnins.2016.00604
M3 - Article
AN - SCOPUS:85011891104
VL - 10
JO - Frontiers in Neuroscience
JF - Frontiers in Neuroscience
SN - 1662-4548
IS - JAN
M1 - 604
ER -