Estimation of cerebral physiology and hemodynamics via near-infrared spectroscopy

Abstract

Near-infrared spectroscopy (NIRS) is a non-invasive optical imaging technique that has rapidly been gaining popularity for study of the brain. Near-infrared spectroscopy measures absorption of light, primarily due to hemoglobin, through an array of light sources and detectors that are coupled to the scalp. Measurements can generally be divided into measurements of baseline physiology (related to total absorption) and measurements of hemodynamic time-series data (related to relative absorption changes). Because light intensity drops off rapidly with depth, NIRS measurements are highly sensitive to extracerebral tissues. Attempts to recover baseline physiology measurements of the brain can be confounded by high sensitivity to the scalp and skull. Time-series measurements contain high contributions of systemic physiology signals, including cardiac, respiratory, and blood pressure waves. Furthermore, measurements over time inevitably introduce artifacts due to subject motion.

The aim of this thesis was to develop improved analysis methods in the context of these NIRS specific confounding factors. The thesis consists of four articles that address specific issues in NIRS data analysis: (i) assessment of common data analysis procedures used to estimate oxygen saturation and hemoglobin content that assume a semi-infinite, homogeneous medium, (ii) testing the feasibility of improving oxygen saturation and hemoglobin measurements using multi-layered models, (iii) development of methods to estimate the general linear model for functional brain imaging that are robust to systemic physiology signals and motion artifacts, and (iv) the extension of (iii) to an adaptive method that is suitable for real-time analysis. Overall, this thesis helps to validate and advance analysis methods for NIRS.