EEG-based control of working memory maintenance using closed loop binaural stimulation


This project, funded by NSF, targets people with mild cognitive impairment (MCI), which is a form of cognitive impairment characterized by dysfunction in memory, language, logical reasoning, and judgement more severe than the typical cognitive decline associated with normal aging. Through this project, we seek to discover noninvasive methods for identifying the neural networks associated with working memory in healthy individuals and in people with MCI, and to improve working memory in both populations using a non-invasive neuromodulation technique in the form of an auditory stimulus. Working memory is selected since its decline is often an early indicator of MCI. We expect this work to improve the quality of life for people with MCI by offering a noninvasive therapeutic intervention based on so-called binaural beats in its future incarnations.
The project seeks to answer the question of 1) whether binaural beats can affect working memory 2) whether they can be used to control working memory circuitry 3) what is the effect of a binaural beat stimulus on key brain regions associated with both auditory processing and working memory using computational and biophysical models of how the brain accomplishes working memory, we study, in simulation. In addition, we develop hardware for implementing real-time control of brain signals using binaural beats. We develop adaptive controllers to use in closed loop control of the brain signals. The brain signals are recorded using electroencephalography (EEG) which noninvasively measures the voltage across the scalp using a series of electrodes. Furthermore, we experimentally test hypotheses driven by numerical studies on the mathematical model. The goal of this work is developing an adaptive controller with the assumption of unknown brain model and identifying EEG outputs which can reliably predict a subject’s ability to perform working memory tasks and to control the subject’s performance using closed loop control of the binaural beat stimulus, both in people who are neurologically healthy or who have MCI.
The frequency response of the brain has been associated with a wide range of brain states and abilities, including concentration, mood, attention, and memory. These spectral correlates emerge from synchronized activity among ensembles of neurons and are now appreciated as supporting system-level information encoding in addition to neural spikes. To access particular neural responses, a variety of interventions target stimulating the brain with electrical signals, magnetic fields, and ultrasounds. However, few existing methods take advantage of the brain’s own structure to actively generate internal oscillatory modulations. Our project aims to use binaural beats, which arise from the brain’s interpretation of two pure tones, with a small frequency mismatch, delivered independently to each ear. The mismatch between these tones is perceived as a so-called beat frequency. The use of binaural beats to entrain certain brain structures has been preliminarily explored and results suggest that this safe and accessible method can be used to modulate behavioral performance. For designing controller, the key idea is that, in the circumstance that the system model is unknown and the state vector is unmeasured, it is not necessary to construct the model dynamics and attempt to estimate the full state if the output (i.e. the measurements) is all we need to control. We introduce an output predictor, capable of predicting the system output using the history of the system input and output stored in autoregressive filtered vectors. Hence, designing an output tracking control for the unknown system is equivalent to constructing a tracking control for the predictor, which is a virtual system whose dynamics and state are known. With this approach, the tracking task can be achieved by designing a tracking controller for a linear time varying system, using one of many approaches existing in the literature. Exponential stability of the prediction error and uniformly ultimate boundedness of the tracking error are proved using Lyapunov's direct method. This project will advance knowledge and state-of-the-art in adaptive predictor-based output feedback control for a class of unknown SISO linear systems. It also supports the hypothesis that the binaural beats create an environment where working memory is improved. This project will be helpful in understanding if binaural beats have the potential to control brain oscillations in a closed control loop in some people without MCI. Improving the working memory for healthy people will have a great impact economically, where industry can benefit from having more efficient people. The results of this work are expected to act as a foundation for closed-loop control studies of noninvasive brain stimulation, demonstrating all aspects of the systems design from hardware integration to tuning parameters, as well as providing a first quantitative assessment of binaural beats and their potential as a stimulus.


Alexander Leonessa

Professor, Mechanical Engineering Department

Virginia Tech, Blacksburg, VA 24060

Nicole Abaid
Associate Professor, Biomedical Engineering and Mechanics
Virginia Tech, Blacksburg, VA 24060

Rosalyn Moran
Assistant Professor
Virginia Tech Carilion Research Institute
Senior Lecturer in Mathematical Neuroscience
Department of Engineering Mathematics
University of Bristol, UK