Overview: Our research covers a variety of topics in computational neuroscience, biomedical signal processing, neural engineering, and digital signal processing. Our recent work is centered around neuroinformatics and neuroengineering. Major research directions are outlined below.
Estimation of dynamic brain networks from functional neuroimaging data
Exploration of connectivity between spatially distributed brain regions is a central research theme in neuroscience. At present there is a rapidly increasing interest in the temporal variations of brain connectivity, which enables the exploration of how brain regions dynamically exchange information to support cognition and behavior and how such a coordination could be affected by neuropsychiatric disorders. This research is aimed to develop connectivity analysis methods to estimate how the human brain is dynamically updated and orchestrated. Further, we aim to investigate the functional significance and clinical relevance of the dynamic brain networks.
Identification of neural correlates of human pain perception
Pain is a subjective first-person unpleasant multidimensional experience. Self-report is the gold standard for the determination of the presence, absence, and intensity of pain in clinical practice. However, self-report of pain fails to be used in some vulnerable populations (e.g., patients with disorders of consciousness), which can lead to various serious clinical problems. Therefore, the availability of a physiology-based assessment of pain would be of great importance in basic and clinical applications. This research is aimed at identifying the cortical activity related to the generation of painful perception.
Pattern recognition and machine learning for brain decoding
Pattern recognition and machine learning techniques are increasingly used to identify brain activation patterns corresponding to external stimuli or cognitive/behavioral responses and to further infer mental states from brain signals. In this study, we aim to develop and apply pattern recognition and machine learning techniques to identify discriminative neural features from high-dimensional neuroimaging data for higher prediction accuracy and better model interpretability. For example, we are interested in cross-subject prediction (the prediction model is trained on a cohort of individuals and applied to another individual), which would be of great clinical value.
Signal processing methods and system design for neural engineering applications
Real-time and accurate identification of brain activity (mainly EEG) underlying sensory and cognitive information processing is central in neural engineering applications, such as brain-computer interface. However, meaningful but weak information is usually buried in a high amount of background noise and non-cortical artifacts, and, therefore, cannot be easily characterized. To address this problem, we aim to develop and implement advanced signal processing methods to obtain robust and reliable detection of underlying mental states from EEG in a real-time and automated manner.
Processing and fusion of multimodal brain imaging signals
The human brain activity can be measured from different perspectives with different imaging technologies. A recent trend in neuroimaging is to simultaneously acquire multimodal signals with complementary properties so that acquired information can be integrated to provide a more complete picture about brain functioning. Here, we aim to develop signal processing and modelling methods for the integration of EEG and fMRI measurements, which can achieve both good spatial resolution provided by fMRI and good temporal resolution provided by EEG.