Brain-computer interfacing (BCI) can read thoughts and recognize brain states from neural signals and translate information to external machines/devices. EEG, as a non-invasive, portable, low-cost modality, can track brain dynamics in high temporal resolution. Yet various challenges, including the issues of signal quality, data variability, and convenience of use, hinder laboratory-based EEG systems towards real-world applications. Our group is developing advanced methods to implement more efficient and effective BCI applications.
Explainable AI (XAI) modeling is a powerful tool for exploring neuroscience. Instead of traditional research methodologies, our group has developed light-weighted, interpretable neural networks not only for high-performance brain decoding, but also for extracting pattern of interest in the brain activities that can shed light on future brain research.
Current neuromodulation techniques (e.g. TMS) adopt fixed protocol without considerations of individual differences in neurological/psychiatric therapies. Our group incorporates computational approaches and neuroscience methods to facilitate close-loop non-invasive, personalized neuromodulation with dynamic parameter tuning according to brain dynamics. With enhanced effectiveness of brain stimulation, we foresee improved clinical outcomes using the dynamic personalized neuromodulation.