Hau-Tieng Wu: New Mathematical Methods for Analyzing Complex Physiological Time Series

ABSTRACT: High-frequency physiological waveform signals, such as ECG, PPG, EEG, etc., provide rich real-time information about the well-being of the patient. However, the analysis of these physiological signals remains a challenge, particularly when the signal is long and nonstationary. In this talk, I will propose a novel approach for the analysis of complex physiological time series. I will demonstrate the approach by providing a solution to the single-channel blind source separation challenge, which is commonly encountered in different clinical setups, like the home-care monitoring system. In addition to introducing the algorithm, the theoretical foundation of the solution based on nonlinear-type time-frequency analysis and differential geometry will be provided, particularly when the signal has complicated statistical features, like time varying amplitude, frequency and non- sinusoidal pattern, and non-stationary noise. Its clinical application to extracting fetal ECG morphology from the single channel maternal trans-abdominal ECG signal for the long-term monitoring purpose and stimulation artifact removal for the EEG signal will be discussed.
BIO: Wu earned his Ph.D. in mathematics from Princeton University in 2011 under the supervision of Professor Ingrid Daubechies. While at Princeton he also worked with Professor Amit Singer. He performed postdoctoral research at Princeton’s Program in Applied and Computational Mathematics in 2011, and in statistics at UC Berkeley in 2012 and in mathematics from Stanford University in 2013. He started his tenure-track position in mathematics at University of Toronto in 2014. He moved to mathematics and statistical science at Duke University in 2017. Before moving to mathematics, he obtained his medical degree from National Yang-Ming University in Taiwan in 2003, and practiced in Veteran General Hospital, Taipei, Taiwan for 1+ years. He is passionate about data science with solid theoretical foundation for medical data analysis. He is not only engaged in analyzing multimodal physiological waveform signal (high dimensional time series) by developing suitable signal processing algorithms, but also in establishing their mathematical and statistical foundation. He cares about not only providing interpretable physiological information for clinical decision making, but also about understanding the dynamical system guiding human physiology and pathophysiology. In addition to collaborating closely with physicians, he also collaborates with hardware engineers to implement the established algorithms toward clinical usage.