UC San Diego

The human sensorimotor cortex contains topologically organized maps of the body and movements, usually referred to as Penfield’s homunculus. The initial homun-cular map was depicted using electrical stimulation in patients undergoing brain surgery. For the last 25 years, noninvasive functional magnetic resonance imaging (fMRI) has allowed researchers to map brain organization and function with high spatial resolution (on the order of millimeters). However, less research has been conducted to model the temporal dynamics of fMRI for its low temporal resolution (on the order of seconds). To map topological organization in visual, auditory, and sensorimotor cortices, the phase-encoded paradigm uses time delays in periodic fMRI signals to encode and decode the representations of different brain regions. Phase-encoded fMRI data are typically ana-lyzed by Fourier transform, and an overall signal-to-noise ratio and a single phase of periodic signals are estimated from the entire time series. However, this method cannot fully characterize the temporal dynamics and stability of periodic signals and phases.

This dissertation includes three fMRI studies that aim to map the cortical repre-sentations of body parts (face and hand) and unravel the spatiotemporal brain dynamics during a reach-to-eat task. I proposed a data processing pipeline that uses independent component analysis to separate periodic noise from periodic signals, and use time-frequency analysis and circular statistics to model the spread of phase angle of fMRI time series and determine the stability of periodic signals. These analyses provide additional statistical measures, not obtainable using conventional linear methods, to validate the observed periodic fMRI signals. These methods are fundamental steps to construct a more accurate and comprehensive functional brain atlas of the human sensorimotor cortex.