Dynamic analysis of resting state fMRI data and its applications

While most resting state connectivity studies assume that resting-state fMRI time series are stationary, there is growing evidence indicating that they are in fact dynamically evolving. This paper describes two pieces of our work related to the resting state dynamics. We assume the resting-state brain to be in quasi-static states with spontaneous switching between them. First, we apply a hidden Markov model to the resting state fMRI data and derive model parameters reflecting the states. With this approach, we identified 9 reproducible states, which resemble resting state networks described in the literature. The second piece of work is the dynamic parcellations of thalamus, leading the state specific parcellations and their merged results, both of which revealed new insights about the thalamic function and connectivity.