Stress mitigation through adapted binaural beats

The use of binaural beats (BB) for neurostimulation is an innovative technique useful for regulating brain activity and relieving stress. The aim of this thesis is to develop a closed-loop neurostimulation protocol to investigate whether the use of BB adapted to the subject’s stress in real-time can improve cognitive performance and modulate stress in healthy individuals. This study evaluate whether the adapted stimulation with BB in the alpha band is more effective compared to stimulation with constant BB and to the control condition without stimulation. The research included the validation of artificial intelligence models to identify stress levels in real time using physiological signals such as electroencephalogram (EEG) and the electrocardiogram (ECG). The experimental protocol required each participant to complete four acquisition sessions, each including three minutes of relaxation and five minutes of mathematical operations with two levels of difficulty. The four experimental settings included a calibration phase to train the regressor for stress recognition, a session with constant BB stimulation, a session with adaptive BB stimulation and a control condition without stimulation. Calibration was always conducted first, while the other sessions were randomized. During the experiment, the subjects listened to relaxing music to isolate them from external noise, with brown noise added in neurostimulation phases to mask BB stimuli, making them inaudible. In the adaptive stimulation phase, the modulation was performed by assessing the subject’s stress state in real time every ten seconds and changing the beat frequency of the stimulation accordingly, while the other phases were conducted in offline mode. EEG and ECG data were recorded using the Enobio 8-channel device and the NIC2 software. The data collected were used to train three different machine learning models. The Random Forest (RF) showed adequate timing for real-time acquisition and good performance in recognizing stress levels. EEG and ECG data from twenty healthy subjects were analyzed to identify significant features discriminating between the stressful and relaxed state, while cognitive abilities were assessed through reaction time (RT), accuracy of responses and the inverse efficiency score (IES). Results indicate that both stimulation conditions improve cognitive performance and reduce stress levels compared to the control condition. Statistical tests showed that the use of adaptive BB had a more pronounced effect, with a significant decrease in RT and an improvement in IES scores compared to both constant BB (p < 0.001) and the control group (p < 0.001). The Fisher Ratio identified standard deviation, range, and maximum value as the most useful metrics for discriminating between relaxation and stress states, with the right frontal and left parietal as the most relevant hemispheres. This study confirms the hypothesis that adaptive stimulation with BB is effective in improving cognitive functions and mitigating stress, highlighting the potential of this technique as a non-invasive resource for mental well-being. The results highlight the variability of individual reactions to BB and the importance of further studies to improve the neurostimulation model, which could include integrating new ECG and EEG metrics, testing different algorithms for stimulation, and using additional physiological and psychological measures to better understand the effects of stimulation on performance and stress mitigation.