Automated Detection of Alzheimer's disease and other neurophysiological applications based on EEG

Abstract

Electrophysiology (EEG) refers to the technique for the acquisition of brain electrical activity using electrodes placed in the scalp. For the past years, hardware and software advancements have promoted the use of wearable EEG systems in multiple research fields such as neuromarketing, sports performance, or the detection of neurological diseases. Simultaneously, artificial intelligence techniques have experienced a remarkable growth, with machine learning approaches widely spreading across research fields. Thereby, researchers in fields like neuroscience and neuroengineering combine EEG feature extraction techniques with machine learning algorithms to study systems based on brain electrical activity for different applications. In this thesis work, we investigated a series of applications combining EEG acquisition, signal processing, and machine learning. Particularly, we focused on three main topics: attention, stress, and Alzheimer’s disease. With respect to attention, we investigated the real-time detection of the attention exerted to mobile visual stimuli, what could help to enhance attention training therapies. For this purpose, we combined EEG and gamification principles to develop a videogame based on brain activity. Regarding stress, we investigated the use of machine learning models to regress the self-perceived stress level during a stressrelax session. This may contribute to improving stress therapies such as those conducted in special needs education schools, where children are often uncapable of communicating verbally. Finally, regarding Alzheimer’s disease, we investigated two main topics focused on delivering new automated detection techniques to assist the clinicians and shorten detection times: (a) an automated approach based on EEG activity and machine learning for the real-time detection of early Alzheimer’s disease, and (b) the detection of cognitive impairment via a computerized cognitive task evaluating visual dynamics. To investigate the mentioned applications, we conducted diverse research studies. As a result, we published a series of articles in scientific journals which we have gathered into a compendium for this thesis. In addition to this compendium, in this document we introduce other valuable works conducted in collaboration with other research teams which have not been published yet. The results originated from the research works presented in this thesis have demonstrated scientific value as indicated by their publication in international scientific journals with high impact factor. Therefore, this work may provide valuable scientific insights which could impact multiple areas in research and society, such as special needs education, stress assessment for professional training or sports performance, and neurological diseases detection.