(664b) Optimal Control of Closed Loop Neural Prostheses

The overall objective of control-theoretic system level analysis of closed loop neural prostheses is to provide the successful transition of Brain-Machine-Interface (BMI) [1]-based neuroprostheses and assistive devices to stable extended use in human subjects suffering from peripheral neuropathies, spinal cord injuries, neuromuscular disorders like Parkinson's and amputations. BMIs provide an interface between the brain and a machine (computer) by interpreting electrophysiological measurement of neural activity in real time, and translating this information to provide actuation of motor tasks such as hand grasping in artificial arms. The interface with the brain is facilitated by invasive measurements employing intracortically implanted micro-electrodes [2]. Controlling such prosthetic devices using BMIs requires an appropriate control architecture design which can provide appropriate input commands to motors for accomplishing a smooth movement of the limb. Past research has focused on decoding neuronal population in rats for the general goal of controlling prosthetic devices. The experiments have actuated, for example, a robotic device [5], in primates for reproducing reaching and grasping movements [6] and in motor impaired human subjects [3, 4]. However, much of this work has utilized implementation of an open loop control for such devices or producing movements. Lack of proper incorporation of feedback information to the system in above studies makes the prosthetic devices prone to error in decoding and targeting the intended action of the neurons. There exist at least three types of feedback paths Natural Sensory Feedback, Surrogate Sensory Feedback, and Neuronal Microstimulation Feedback, that influence the functionality of the prosthetic device by continuously changing the output from the brain. Proper incorporation of these feedback paths, by closing the loop, is necessary for development of normal behaving prosthetic devices.

This presentation introduces a control-theoretic framework that considers various interacting systems and feedback paths for designing stable neural prosthetic devices that are comparable to normal limb movements. These interacting systems consist of a ?brain model block?, a ?decoder block?, an ?arm controller block? with ?arm model block?, and a feedback loop with an ?encoder block? to the brain block. The brain model block computes spike patterns for motor cortex neurons using a multi-neuron model proposed by Izhikovich [7]. These spike patterns are compared with experimentally obtained physiological data from a primate corresponding to specific movements tasks to decode the information for motor related tasks such as set points for arm controller (force or position coordinates). With these set points, the arm controller block computes inputs to actuate arm movements in a predictive sense. A Linear Model Predictive Control algorithm is used to compute inputs such as required torque for arm movements. Outputs from the arm model block are fed to the controller as well as to the encoder block. This encoder block uses an encoding algorithm (inverse to the decoder model) to compute stimuli which are then fed to the brain model. The most difficult and challenging task in designing above mentioned control-theoretic framework for a closed loop neural prosthesis is to develop the models for each of the above mentioned blocks with feedback information. For instant, the feedback information from the arm block to the neuron model continuously modifies the behavior of spikes patterns which are coming out of the neuron model block. To understand this effect of feedback on neuronal model spikes patterns, we study the effect of different patterns of action potential spikes stimuli on the neuronal model in an open loop framework. From this information, and from experimentally obtained electrophysiological data for a primate, we fit the neuronal model parameters to match the model output with experimentally obtained data. The model parameters are continuously updated at every instant of time using the feedback information and encoder model based on experimental results. The visual feedback information from the prosthetic device is encoded into action potential spikes using models stated in literature. We use dynamical models for finger movements (flexion and extension) to represent the arm block and study the finger movements within this closed loop control framework. Results from this study may promise the development of a stable neural prosthetic that can be used for human subjects.

References

[1] M. A. L. Nicolelis. Brain-machine interfaces to restore motor function and probe neural circuits. Nature Reviews Neuroscience, 4:417-422, May 2003.

[2] J. C. Sanchez, J. M. Carmena, M. A. Lebedev, M. A. L. Nicolelis, J. G. Harris, and J. C. Principe. Ascertaining the importance of neurons to develop better Brain-Machine Interfaces. IEEE Transactions on Biomedical Engineering, 51(6):943-953, June 2004.

[3] L. R. Hochberg, M. D. Serruya, G. M. Friehs, J. A. Mukand, M. Saleh, A. H. Caplan, A. Branner, D. Chen, R. D. Penn, and J. P. Donoghue. Neuronal ensemble control of prosthetic devices by a human with tetraplegia. Nature, 442:164-171, July 2006.

[4] P. R. Kennedy, M. T. Kirby, M. M. Moore, B. King, and A. Mallory. Computer control using human intracortical local field potentials. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 12(3):339-344, September 2004.

[5] J. K. Chapin, K. A. Moxin, R. S. Markowitz, and M. A. L. Nicolelis. Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex. Nature Neuroscience, 2(7):664-670, July 1999.

[6] J. Wessberg, C. R. Stambaugh, J. D. Kralik, P. D. Beck, M. Laubach, J. K. Chapin, J. Kim, Mandayam A. Srinivasan, S. James Biggs, and M. A. L. Nicolelis. Real-time prediction of hand trajectory by ensembles of cortical neurons in primates. Nature, 408:361-365, November 16, 2000.

[7] E. M. Izhikevich. Simple model of spiking neurons. IEEE Transactions on Neural Networks, 14(6):1569-1572, November 2003.

Acknowledgment:

We acknowledge Dr. Marc Schieber, U. of Rochester Medical Center for providing us with primate data on intracortical recordings.

Financial support from the US National Science Foundation, Cyber Enabled Discovery and Innovation (CDI) program, is gratefully acknowledged.