Speaker: Xiaoping Hu,Ph.D.(The University of California, Riverside)
Time: 13:00-14:30, June 18, 2019
Venue: #1113, Wang Kezhen Building
Abstract: Resting state functional MRI (rsfMRI) is becoming widely used for studying the brain. While in the early year of rsfMRI, the fMRI signals were assumed to be stationary, it has been recognized for a decade that the they are not stationary and contain dynamic features that are of great value in probing the resting brain activity and connectivity. Although a fully dynamic system is difficult to describe and characterize, it is now well accepted that the resting brain could be considered as hopping between a number of well-defined states. In this talk, I will first describe our approach for characterization the brain states and their dynamics using a hidden Markov model. This approach was able to identify brain states and their temporal features. Second, recognizing the notation of brain states, we developed a state dependent parcellation scheme for parcellation the thalamus, a structure that is known to play an important role of relay and control. Finally, I will demonstrate how the rsfMRI dynamic features can augment rsfMRI spatial features for individual identification using machine learning. These features not only led to robust individual identification but also shed lights on individual variability in spatiotemporal features in a large population. Overall, this talk will demonstrate that dynamic features in resting state fMRI data can be characterized and exploited in novel ways.
Host: Prof. Fang Fang