The cerebellum, a critical part of the brain, is mainly responsible for motor coordination and sensorimotor learning. It enables accurate and precise movement which requires precision on a millisecond time scale through various forms of learning. Damage to the cerebellum, caused by different brain diseases and injuries, can result in motor and cognitive impairment, like ataxia, which severely impacts the patient’s daily life. There is still no complete cure to cerebellar neurological disorders, due to the unclear working mechanism of the cerebellum and limited therapeutic methods. An explicit neural circuit is the foundation for developing new therapeutic methods, such as deep brain stimulation therapy and neuroprosthetics. Therefore, in this thesis, we aim to further investigate the neural circuit of the cerebellum and develop more effective neuroprosthetic systems by implementing the neural circuit to computational simulation.

In this thesis, based on the primary neural circuit of cerebellar associative motor learning, we reveal that a new feedback projection, from the interpositus nucleus (Int) to the pontine nucleus (PN), plays an important role in cerebellum-mediated motor learning. Meanwhile, we simulated the neural circuit with a spiking neural network (SNN) model and established a real-time neuroprosthetic system by implementing the simulation model to a field-programmable gate array (FPGA).

In the first part of the study, we took advantage of the delay eyeblink conditioning (dEBC) model, which is a classical model for cerebellar associative motor learning and computational simulation, to study the mechanism of the Int-to-PN projection on cerebellum-mediated motor learning. The experiments were conducted from the perspectives of neurotropic virus tracing and neural anatomy, animal behavior and the characteristics of neuronal firings. Results show the following: (1) the Int-to-PN projections are glutamatergic and distribute mainly in the caudal middle part of the PN. (2) inhibition or excitation of the axonal terminals of the Int-to-PN projections decrease or increase, respectively, the rate of establishment of dEBC, and the conditioned response (CR) window is the specific time window of efficiency; (3) the excitatory Int-to-PN projection plays a facilitating role in associative motor learning by strengthening the conditioned stimulus (CS) input signals to the cerebellar cortex in two ways: a) by magnifying and maintaining the CS inputs by modulating CS&CR-relay neurons in the PN; b) by transferring the corollary discharge of output signals of the Int to the cerebellar cortex via CR-relay neurons in the PN. These results clarify the critical role of the Int-to-PN projection in cerebellum-mediated motor learning, completing the neural circuit of the cerebellum by adding the feedback regulation further.

In the second part of the study, we developed a real-time cerebellar neuroprosthetic system using an SNN model that simulates synaptic plasticity to learn dEBC. The simulation model was then interfaced with anesthetized rats through hardware implementation of an FPGA. This rat-FPGA hybrid system was able to process neuronal spikes in real-time with an embedded cerebellum model of ~10,000 neurons and reproduce learning of dEBC at different inter-stimulus intervals in the animals. Our results validate that the system performance is physiologically relevant at both neural (firing pattern) level and behavioral (eyeblink pattern) level. This integrated system provides sufficient computation power for mimicking the cerebellar circuit in real-time. Further, this system can interact with the biological system naturally at the spike level and it can be generalized to other neural components (neuron types and plasticity) and neural functions for potential neuroprosthetic applications.

The results in this thesis have significance in clarifying brain mechanisms for cerebellar learning and in facilitating a deep understanding of the general principle of learning and memory. The results are also important in the development of more reliable and robust computer models, brain-machine interfaces, and neuroprosthetics for clinical therapy in the future.