In spite of the worldwide financial and research efforts made, the pathophysiological mechanism at the basis of Alzheimer's disease (AD) is still poorly understood. Previous studies using electroencephalography (EEG) have focused on the slowing of oscillatory brain rhythms, coupled with complexity reduction of the corresponding time-series and their enhanced compressibility. These analyses have been typically carried out on single channels. However, limited investigations have focused on the possibility yielded by computational intelligence methodologies and novel machine learning approaches applied to multichannel schemes. The study at screening level on EEG recordings of subjects at risk could be useful to highlight the emergence of underlying AD progression (or at least support any further clinical investigation). In this work, the representational power of Deep Learning on Convolutional Neural Networks (CNN) is exploited to generate suitable sets of features that are then able to classify EEG patterns of AD from a prodromal version of dementia (Mild Cognitive Impairment, MCI) and from age-matched Healthy Controls (HC). The processing system here used enforces a series of convolutional-subsampling layers in order to derive a multivariate assembly of latent, novel patterns, finally used to categorize sets of EEG from different classes of subjects. The final processor here proposed is able to reach an averaged 80% of correct classification with good performance on both sensitivity and specificity by using a Multilayered Feedforward Perceptron (MLP) of the standard type as a final block of the procedure.

Alzheimer's Disease (AD) is a neurodegenerative disorder that causes a loss of connections between neurons. The goal of the present work was to construct a complex network model of the brain-electrical activity using high-density EEG (HD-EEG) recordings and to compare the network organization in AD, Mild Cognitive Impaired (MCI) and healthy control (CNT) subjects. The HD-EEG of 16 AD, 16 MCI and 12 CNT were recorded during an eye-closed resting state. The Permutation Disalignment Index (PDI) was used to describe the dissimilarity between EEG signals and to construct the connection matrices of the network model. The three groups were found to have significantly different ( $p\lt0.001$ ) Characteristic Path Length (Lambda), average Clustering Coefficient (CC) and Global Efficiency (GE). This is the first time that HD-EEG signals of AD, MCI and CNT have been compared and that PDI has been used to discriminate between the three groups. Considered the large amount of data originating from the HD-EEG acquisition, compared to standard EEG, the aim of the present work was also to assess that compression did not alter the results of the complex network analysis. Compressive sensing was adopted to compress and reconstruct the HD-EEG signals with minimal information loss, achieving an average Structural SIMilarity index (SSIM) of 0.954 (AD), 0.957 (MCI) and 0.959 (CNT). When applied to the reconstructed HD-EEG, the complex network analysis provided a substantially unaltered performance, compared to the analysis of the original signals. Indeed, Lambda, CC, GE and of the three groups were still significantly different ( $p\lt0.001$ ).