When learning new skills, experience with previously-learned skills can facilitate faster learning by constraining behavioral exploration and shaping, a concept known as ?structural learning?. The motor cortex plays an essential role in learning new skills, and its initially variable activity is shaped and consolidated over learning. However, how previous experience modulates exploration and shaping of cortical network activity to facilitate new skill learning is not well understood. When the brain learns to control a brain-machine interface (BMI), cortical network activity exploration and shaping is broad (high-dimensional) in BMI-nave subjects and constrained (low-dimensional) in BMI-experienced subjects, suggesting the following hypothesis. Hypothesis: Previous experience facilitates faster learning of new, related skills by constraining how motor cortical network activity is explored and shaped, effectively reducing the number of neural parameters to learn. The hypothesis? prediction is that during faster learning of related skills, neural dimensionality will be decreased and aligned with previously learned neural patterns. This project tests the prediction by leveraging novel closed-loop paradigms, chronic large-scale 2-photon calcium imaging, high-dimensional data analysis, and holographic optogenetic stimulation to study and manipulate the neural basis of structural skill learning. First, the correspondence between structural learning of muscle patterns and cortical network activity exploration and shaping will be studied using large-scale 2-photon calcium imaging. Second, to causally link neural variance to learning neural patterns, a high-performance, calcium imaging-based BMI will be developed, and the relationship between structural neuroprosthetic learning and neural exploration and shaping will be analyzed. Finally, the structure of cortical network activity will be artificially shaped using holographic optogenetic stimulation and tested on neuroprosthetic skill learning. The long-term objective of this proposal integrates several core goals of the BRAIN initiative. The proposal will produce a dynamic picture of the learning brain and demonstrate causality using BMIs and holographic optogenetic stimulation. This work?s outcome will contribute conceptual principles underlying skill learning and memory and guide the design of BMI systems to restore movement and assist learning.
Aim 1 : Investigate the relationship between structural motor learning and cortical network activity exploration and shaping using a novel motor task and large-scale 2-photon calcium imaging.
Aim 2 : Investigate the relationship between structural neuroprosthetic learning and cortical network activity exploration and shaping using a high-performance, calcium imaging-based BMI.
Aim 3 : Artificially shape structure of cortical network activity using closed-loop holographic optogenetic stimulation and test effect on neuroprosthetic learning.
Our motor memory lets us execute skills effortlessly by triggering learned patterns of neural activity in our motor cortex, but it also enables us to more quickly learn skills related to those previously-learned. Understanding how motor memory changes the way new motor cortex activity patterns are learned to enable faster skill learning is a fundamental problem concerning our brain?s motor system, memory, and learning. This knowledge will be useful to understand motor disorders and design better brain-machine interfaces which decode neural activity to restore movement and communication to paralyzed patients and assist learning new neural activity patterns.
