Closed-Loop Regulation of Internal Brain States using Wearable Brain Machine Interface Architectures with Real-World Experimental Implementation

The brain is a control system with a strong impact on all human functions. Inspired by the recent advances in wearable technologies, we design wearable-machine interface (WMI) architectures for controlling brain responses. The WMI architecture encompasses collecting physiological data using wearable devices, inferring neural stimuli underlying pulsatile signals, estimating an unobserved state based on the underlying stimuli, designing the control, and closing the loop. In this thesis, we design WMI architectures for regulating human’s cognitive stress state and controlling energy levels in patients with hypercortisolism. Hypercortisolism, which corresponds to the excessive levels of cortisol hormone, is associated with tiredness and fatigue during the day and disturbed sleep at night. Automating the use of medications that are effective by either elevating or lowering the energy levels might help patients with hypercortisolism to experience more balanced energy cycles required for their daily activities and better sleep patterns at night. Keeping cognitive stress at a healthy range can improve the overall quality of life by helping the subjects to decrease their high levels of arousal to relax them and elevate their low levels of arousal to increase the engagement. Skin conductance data provides us with valuable information regarding one's cognitive stress-related state. We propose to use this physiological data collected via wearable devices to infer individuals' arousal state. In the first part of this research, we simulate multi-day cortisol profile data for multiple subjects both in healthy conditions and with Cushing's disease. Then, we present a state-space model to relate an internal hidden cognitive energy state to subject's cortisol secretion patterns. Particularly, we consider circadian upper and lower bound envelopes on cortisol levels, and timings of hypothalamic pulsatile activity underlying cortisol secretions as continuous and binary observations, respectively. By estimating the hidden energy state and incorporating the simulated hypothetical medication dynamics, we design a knowledge-based control system and close the loop. In the second part of this research, we design a simulation environment to control a cognitive stress-related state in a closed-loop manner. Hence, using the state-space approach, we relate internal cognitive stress state to the changes in skin conductance. Then, we estimate the hidden stress state and close the loop by designing a fuzzy controller. Next, we propose supervised control architectures to enhance the closed-loop performance in cognitive stress regulation. To further enhance the closed-loop design, we consider adaptive and robust control systems to handle model uncertainty and additional disturbance input. Finally, we design and perform multiple human-subject experiments to further explore safe actuation to regulate internal hidden brain states in real-world. In these novel experiments, we employ wearable technologies and publish data sets that could be further investigated to model the dynamics of proposed safe actuation. These studies are the first steps toward the goal of treating similar mental and hormone-related disorders in real-world situations. Analyzing the human subjects’ responses to the effective safe actuation would further enhance the efficiency of proposed approaches and lead us to a practical automated personalized closed-loop architecture.