UC Merced

Resting-state functional connectivity has revolutionized neuroimaging in the past 20 years and helped us gain insights into the functional organization of the human brain. Although most studies use functional magnetic resonance imaging (fMRI) for studying functional connectivity, functional near-infrared spectroscopy (fNIRS) has become increasingly popular due to its advantages over fMRI. However, fNIRS data is often corrupted by systemic physiology, extracerebral signal contamination, and motion artifacts which limit its widespread use. These noise sources in resting-state fNIRS data often violate the assumptions of the linear models used to model functional connectivity and may lead to incorrect inferences about statistical significance. This dissertation aims to examine the impact of these noise sources in resting-state fNIRS data and proposes preprocessing strategies and connectivity models that ameliorate their effects and improve the statistical validity of functional connectivity analysis. Chapter 2 examines multiple analysis pipelines using simulated resting-state fNIRS data to find the best strategies to correct for the effects of global systemic physiology on resting-state functional connectivity. Our results indicate that pre-filtering using principal components extracted from short-separation fNIRS channels as part of a partial correlation model was most effective in reducing spurious correlations due to shared systemic physiology. Given the high temporal resolution of fNIRS, modeling the lagged relationships between brain regions is crucial in understanding the flow of information over time. Hence, we also explored methods to model these lagged relationships while controlling for the effects of noise using multivariate Granger causality analysis. Chapter 3 surveys some strategies and techniques that have been proposed to correct for temporal autocorrelation, systemic physiology, and motion artifacts. Together, these two chapters highlight the challenges for resting-state functional connectivity analysis with fNIRS, emphasizing the need for better strategies to remove the effects of noise in the service of obtaining statistically valid inferences about the relationships between cortical regions.