Combining EEG signal processing with supervised methods for Alzheimer’s patients classification

• DOI: 10.1186/s12911-018-0613-y
• Publication Date: 2018-05-31T12:23:14Z
• Authors: Giulia Fiscon; Emanuel Weitschek; Alessio Cialini; Giovanni Felici; Paola Bertolazzi; Simona De Salvo; Alessia Bramanti; Placido Bramanti; Maria Cristina De...
• Keywords: Aged, 80 and over; Male; Computer applications to medicine. Medical informatics; R858-859.7; Electroencephalography; Signal Processing, Computer-Assisted; 02 engineering and technology; Alzheimer's disease; Middle Aged; Classification; Electroencephalography signals; 3. Good health; AD; machine learning; EEG signals; wavelet analysis; fourier analysis; Alzheimer Disease; 0202 electrical engineering, electronic engineering, information engineering; Feature extraction; Humans; Cognitive Dysfunction; Female; Alzheimer’s disease; Alzheimer's disease; Classification; Electroencephalography signals; Feature extraction; Aged; Aged, 80 and over; Alzheimer Disease; Classification; Cognitive Dysfunction; Electroencephalography; Female; Humans; Male; Middle Aged; Signal Processing, Computer-Assisted; Research Article; Aged

Alzheimer's Disease (AD) is a neurodegenaritive disorder characterized by progressive dementia, for which actually no cure known. An early detection of patients affected AD can be obtained analyzing their electroencephalography (EEG) signals, show reduction the complexity, perturbation synchrony, and slowing down rhythms. In this work, we apply procedure that exploits feature extraction classification techniques to EEG whose aim distinguish patient from ones Mild Cognitive Impairment (MCI) healthy control (HC) samples. Specifically, perform time-frequency analysis applying both Fourier Wavelet Transforms on 109 samples belonging AD, MCI, HC classes. The designed with following steps: (i) preprocessing signals; (ii) means Discrete Transforms; (iii) tree-based supervised methods. By our procedure, are able extract reliable human-interpretable models allow automatically assign into class. particular, exploiting achieve 83%, 92%, 79% accuracy when dealing vs MCI problems, respectively. Finally, comparing performances methods, find out Wavelets outperforms Fourier. Hence, suggest it in combination methods automatic based signals aiding medical diagnosis dementia.