Creation of a VR Environment for the Simulation of the Visual Defect in Hemianoptic Patients and Analysis of Alpha Waves in EEG

Virtual Reality enables immersion in computer-generated three-dimensional environments, allowing interaction through devices like headset, headphones and controllers. Many of the realistic and engaging experiences made possible by this technology have applications in medicine and healthcare. Virtual Reality changes the relationship between reality and virtuality in medicine, allowing researchers to use VR environments to conduct more realistic experiments and clinical studies. This drives scientific progress and contributes to the development of new treatments and therapies. Hemianopsia is a condition that leads to a visual deficit characterized by the loss of half of the visual field. This impairment is caused by lesions in the optic chiasm due to various pathological conditions or by post-chiasmatic lesions in the neural pathways that follow the optic chiasm. The objective of this thesis is to compare the EEG signal in participants with and without a virtually simulated visual defect while navigating in appropriately developed VR environments. Using Unity and Blender software, three VR environments were created to which the visual defect was simulated: a commercial scenario, a domestic scenario and a driving scenario. The experiment was carried out in the Polytechnic of Turin’s 3D Lab where 60 participants wore an EEG headset (EMOTIV Epoc X) and a wrist sensor for measuring skin conductance and heart rate (Shimmer3 GSR). Additionally, eye tracking was recorded using the Oculus Meta Quest Pro headset. During the experimentation of each scenario, participants were asked to perform tasks specifically chosen to simulate the actions that hemianoptic patients have difficulty completing. At the end of the experiment, candidates were asked to complete a questionnaire where they evaluated the sensations experienced during the three scenarios in terms of Arousal, Dominance, Valence, and Stress with a score from 1 to 5. A subsequent analysis was conducted on different biological parameters, and this thesis project is focused on the EEG signal analysis. Parameters such as Arousal, Dominance, Valence, and Stress were extracted from the EEG signal waves. These indicators were classified based on the answers given by the candidates in the questionnaire using four types of Machine Learning algorithms: kNN, Random Forest, XGBoost, and Support Vector Machine. The results were compared by evaluating and confronting some performance metrics such as the accuracy parameter. The objective was to find a correspondence between the questionnaire answers and the indicators extracted from the EEG signal, which was confirmed by the excellent results achieved. In addition, a binary classification was performed using the same classification algorithms: this time the object of classification was the alpha-band power recorded by some electrodes for each VR environment and classified according to the presence of the visual defect. Finally, a statistical analysis of the EEG signal was conducted, comparing the alpha band power recorded by specific electrodes in subjects who performed the experiment with and without the visual impairment simulation. The findings revealed an increase in alpha band power in subjects who performed the experiment without the visual impairment, supporting the knowledge that alpha waves can be considered an electrophysiological imprint of the visual system's functioning.