An integrated approach based on EEG signals processing combined with supervised methods to classify Alzheimer's disease patients

Alzheimer's disease (AD) is the most widespread form of dementia, for which actually no cure is known [1]. Furthermore, a particular disease called Mild Cognitive Impairment (MCI) affects patients that suffer from some isolated cognitive deficit due to which they could develop AD [2]. Diagnosing MCI and mild AD is hard because most symptoms are often ascribed to normal consequences of aging. Nowadays, the diagnosis requires a combination of physical, neurological, and neuropsychological evaluations, and a variety of other diagnostic tests including imaging techniques. In particular, Electroencephalography (EEG) appears as a non-invasive and repeatable technique to diagnose brain abnormalities [3]. Different studies have shown that AD has (at least) three major effects on EEG signals: enhanced complexity, slowing of signals, and perturbations in EEG synchrony. The open challenge is to perform clinical studies in order to shed light on biological and medical questions related to AD and MCI [4]. Despite of technological advances, the analysis of EEG continues to be carried out by experts, who are subject to laborious interpretation of the spectrum. Computational methods may lead to a quantitative analysis of these signals and hence to characterize EEG time series [5]. In this work, we propose an integrated feature extraction and classification method for EEG signals, whose aim is to distinguish patient affected by AD from the ones affected by MCI and healthy control (HC) samples [6]. Specifically, we perform a time-frequency analysis by applying both the Fourier and Wavelet Transforms on 109 samples belonging to AD, MCI, and HC classes. The classification procedure is designed with the following steps: (i) pre-processing of EEG signals; (ii) feature extraction by means of the Discrete Fourier and Wavelet Transforms; and (iii) classification with tree-based supervised methods. By applying our procedure, we are able to extract reliable human-interpretable classification models that allow to automatically assign the patients into their belonging class. In particular, by exploiting a Wavelet feature extraction we achieve 83%, 92%, and 79% of accuracy when dealing with HC vs AD, HC vs MCI, and MCI vs AD classification problems, respectively. Finally, by comparing the classification performances with both feature extraction methods, we find out that Wavelets analysis outperforms Fourier. Hence, we suggest it in combination with supervised methods for automatic patients classification based on their EEG signals for aiding the medical diagnosis of dementia. Processed data from our study are available at ftp://bioinformatics.iasi.cnr.it/public/EEG/.