UC Berkeley

Abstract

Cortical and Striatal Circuits for Learning Adaptive Behaviors and Wireless Ultrasonic Implants for Interfacing with the Nervous System

by

Ryan M Neely

Doctor of Philosophy in Neuroscience

University of California, Berkeley

Professor Jose M. Carmena, Chair

Brain and nerve interface systems have shown early promise for alleviating a wide range of debilitating conditions. In the field of brain-machine interfaces (BMI), movement kinematics have been decoded from cortical neurons and used as a control signal for prosthetic devices. In the periphery, recent insights into the connection between nerves and organ systems has sparked new interest in the therapeutic potential of accessing and altering activity in peripheral nerves. Investigating fundamental mechanisms through which networks of neurons coordinate to produce adaptive responses can inform the design of next-generation nervous system interfaces. However, technological challenges must also be addressed before these systems are ready for widespread clinical adoption.

Using a brain-machine interface paradigm, we trained rodents to volitionally modulate the activity of primary visual cortex (V1) neurons. This approach allowed us to observe the instrumental learning process in cortical networks directly, and define the relationship between neural activity and behavioral outcomes. Learning occurred in the absence of visual input, suggesting that modulations were internally driven. Similar to demonstrations of instrumental learning in other cortical areas, learning in V1 engaged and required activity in the striatum, suggesting that cortico-striatal circuits are an essential component for behaviorally-relevant adaptation of cortical outputs. Next, we investigated how factors affecting behavioral choice are represented by striatal neurons as rodents performed a two-alternative probabilistic switching task. We found a rich encoding of task parameters in the dorsomedial striatum both at the level of single neurons and neural populations. We observed activity related to animals’ confidence in the current state of the task, and found that confidence levels modulated the strength and timing of signals predicting behavioral choice. Finally, we sought to address the limitations of current methods for interfacing with the nervous system. We designed, built, and tested mm-scale wireless implants for recording electrical activity in peripheral nerves and muscles. This system, called neural dust, utilized ultrasonic backscatter as a scalable means for powering and communicating with miniaturized devices implanted deep in tissue. We showed that this system is capable of recording electroneurogram and electromyogram activity with high fidelity in living animals.