A new view of motor cortical function
Abstract:
Given that the motor cortex operates as a node in a complex network, how can its function be studied? Because the arm has redundant muscle activation, moves with multiple degrees-of-freedom governed by equations of motion that are long and complicated, one might expect that behavior variables encoded in neural activity would be difficult to recognize. However, when motor cortical activity is recorded during arm movements, it is clear that the instantaneous direction of the arm is easily extracted from neural activity. Many neurons are modulated by this directionality and because this signal is widely distributed, extraction algorithms based on the firing rates of different neurons can be used to accurately decode this direction. The extraction of arm velocity signals from the brain was the first example of how the use of population decoding could be used to monitor behavior-related information.
Viewing motor cortical activity as a way to monitor information flow in a network differs from conventional neuroscience approaches based on reductionism, discrete locationism, and circuit definition. Instead of assuming that a neuron’s activity has a clear causal role in behavior (e.g. muscle contraction), we can gain a description of network functionality through computational approaches based on ‘hidden’ or ‘latent’ states. These approaches, afforded by new technology to reliably record the activity patterns of many neurons simultaneously, are based on the idea that variables measured during an experiment (e.g. firing rate, movement direction, muscle force) are observations of the time-varying control process that governs movement generation. This process is dynamic- it is both autonomous, operating by intrinsic elements and driven by input. Changes in a system’s intrinsic operation, or ‘state,’ can be recognized by changes in the relation of the observed variables to each other (e.g. a change in the tuning function between direction and firing rate). Input to the system can be found by characterizing the correlational between the firing rate patterns of many simultaneously recorded neurons (e.g. through dimensionality reduction to find common drivers).
We have employed this approach in our work and have results that describe motor control processing from this new point of view. Latent drivers and hidden states, once recognized, can be related to behavioral output and provide some new insight into the overall operation of the motor system.
Bio
Dr. Schwartz received his Ph.D. from the University of Minnesota in 1984 with a thesis entitled “Activity in the Deep Cerebellar Nuclei During Normal and Perturbed Locomotion”. He then went on to a postdoctoral fellowship at the Johns Hopkins School of Medicine where he worked with Dr. Apostolos Georgopoulos, who was developing the concept of directional tuning and population-based movement representation in the motor cortex. While there, Schwartz was instrumental in developing the basis for three-dimensional trajectory representation in the motor cortex. In 1988, Dr. Schwartz began his independent research career at the Barrow Neurological Institute in Phoenix. There, he developed a paradigm to explore the continuous cortical signals generated throughout volitional arm movements. This was done using monkeys trained to draw shapes while recording single-cell activity from their motor cortices. After developing the ability to capture a high fidelity representation of movement intention from the motor cortex, Schwartz teamed up with engineering colleagues at Arizona State University to develop cortical neural prosthetics. The work has progressed to the point that human subjects can now use these recorded signals to control motorized arm and hand prosthetics to feed themselves and perform other tasks of daily living. Schwartz moved from the Barrow Neurological Institute to the Neurosciences Institute in San Diego in 1995 and then to the University of Pittsburgh in 2002. In addition to the prosthetics work, he has continued to utilize the neural trajectory representation to better understand the transformation from intended to actual movement using motor illusions in a virtual reality environment.