Modeling the local field potential in the medial prefrontal cortex in reward-guided learning

Abstract

Brain-machine interface (BMI) technology has made significant progress in enabling people with motor disabilities to control prosthetic limbs. Reinforcement learning (RL) has been adopted in BMIs for training decoders using reward information. Internal rewards, reflected in neural activity in the medial prefrontal cortex (mPFC), can be used for autonomous updates in RL-BMIs. However, relying on spike signals to extract reward information from the mPFC has limitations and will not be available, especially in long-term BMI implants. Local field potentials (LFPs) provide information about neural ensembles and have been proposed as an alternative long-term data source to overcome these limitations. In this thesis, I propose to examine local field potential neuromodulations in mPFC to represent reward information and relate them to synchronized neuronal activities by implementing a data-driven marked-point process (MPP) methodology. We identify synchronized spike probability from the transient events with large power in the LFP broad high frequency(bhf) band (200-400Hz) in the mPFC of rats performing a lever-press task. Compared with extracting local field potential features from the binned spectrogram power, our approach improves 32.42% on average in the peak signal-to-noise ratio (PSNR) between neuron pairs over our data segments. This study contributes to the long-term use of internal reward representation in the mPFC and advances the development of autonomous BMI decoders.

Date of Award	2024
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Yiwen WANG (Supervisor)