|M.Sc Student||Benyamini Miri|
|Subject||Optimal Feedback Control Framework for|
Investigating Neural Modulation during BMI
|Department||Department of Mechanical Engineering||Supervisor||PROF. Miriam Zacksenhouse|
|Full Thesis text|
Brain-Machine Interfaces (BMIs) have been developed to provide a direct communication link between the brain and external devices, bypassing the remaining, potentially injured neuro-muscular system. BMIs are based on the observation that during the execution of reaching movements the neural activity is modulated by several motor parameters including the hand position, velocity, speed, and force. The signals decoded from the neural activity from an ensemble of neurons, are used to control external devices to perform reaching movements.
BMIs also provide a unique window into information representation and processing in the Brain. This work focuses on developing optimal control framework for modeling motor control and corresponding neural representations to explain specific phenomena related to the effect of switching to brain control on the variance of neural activity. Specifically, it has been shown that the overall modulations of the firing rate of cortical neurons increase after monkeys started operating the BMI. In contrast, neural modulations that are correlated with the movement kinematics remain relatively unchanged.
Following recent hypotheses in motor control, we assume that the brain implements optimal feedback control. The control of both the hand and the cursor were modelled using linear-quadratic-Gaussian (LQG) control method. The internally estimated state and the control signal are encoded into neural activity under the assumption that the spike trains are realizations of inhomogeneous Poisson processes.
Switching our model from pole control to brain control results in the same phenomena observed in the BMI experiments: The overall modulations increased while the kinematic related modulations remained unchanged. In order to investigate the source of these phenomena, I investigated the effect of higher process noise, either actual or also internal, on neural modulations during normal reaching movements. I demonstrate that increasing the process noise has similar effects: a dramatic increase of overall modulations with minor change in kinematic-related modulations. Furthermore, I trace these phenomena to increasing variance of the encoded motor variables, i.e., the control signal and the estimated state.
I conclude that the increase in overall modulations may reflects changes in the variance of the encoded variables that are due to changes in the process noise and not due to coding additional variables or increasing variance of the executed movement, which would have affected kinematic-related modulations too.