A Normative, Spiking Network Model for Neuro-motor Control and Movement Disorder

What are the objectives, control architectures and encoded information that govern our everyday volitional movements? For decades, behavioral experimentalists have carefully analyzed human reaching movements under a multitude of conditions and stimuli to understand the hidden control objectives of the brain. Similarly, electrophysiologists have recorded neural firing patterns during reaching movements to understand what individual neurons encode, how they convey information as a population and how they influence the downstream muscle activity. Oftentimes, these results are studied in parallel rather than in conjunction. With our normative approach, we allow hypotheses to prescribe the rules of constructing mathematical models of neuromotor control. These initial hypotheses of the brain’s control architecture have led to scientifically grounded predictions of motor behavior and neural firing patterns. To implement this method, the limb was modeled as a dynamical system that receives its control input from a collection of probabilistically spiking artificial neural networks. These networks are trained to implement hypothesized control schemes in order to move the limb to the target state. The output motor behavior and spiking patterns can then be compared with the observations found in prior studies. Using this method, we have trained models of neural “controllers” that exhibit commonly found phenomena in motor and neuron behavior such as: asymmetric and bell-shaped velocity profiles, directionally tuned firing rates, population vectors, and low dimensional dynamics. We extend this approach to other features of control (state estimators and forward models) to predict neural firing patterns for these features. Additionally, these models are used to test how neuron failure can lead to frequently seen motor behavior pathologies. Here, we show that a normative approach to the study of motor behavior may help accelerate research in this field and bridge the gap between behaviorists and electrophysiologists.