dc.description.abstract | Cerebrovascular reactivity (CVR) is a measure of dilation capacity of
cerebral vasculature. It is an important biomarker for the vascular
functionality and integrity, and may have clinical indications in stroke,
atherosclerosis, Moyamoya disease, multiple sclerosis, brain tumor, and
other neurological disorders. The most commonly used approach to
measure CVR is by applying a physiological maneuver to alter the arterial
carbon dioxide (CO2) concentration (e.g. inhaling a small amount of CO2
which is a potent vasodilator), while continuously acquiring BOLD MR
images. However, the current method suffers from several limitations
related to specificity, sensitivity, and physiological modeling of the
measured signal. The goal of my thesis study is to improve on these aspects
and ultimately provide a clinically-ready CVR imaging procedure that could
v
be immediately translational. The proposal’s goals have been accomplished
through the following specific aims:
Aim 1: Improve the specificity of CVR signal by optimization of
imaging protocol.
Although positive CVR, i.e. increased BOLD signal with CO2
inhalation, is expected in healthy brain, recently the presence of negative
CVR has been reported using the current CVR imaging protocol, which can
potentially compromise the interpretability of CVR data in clinical
applications. In Aim 1, we performed simulation and experimental studies
to provide a mechanistic understanding of this observation and showed that
the negative CVR reported in the literature is an artifactual signal due to
improper selection of imaging parameter. We further re-optimize the BOLD
imaging parameters such that negative CVR is no longer present.
Aim 2: Improve the sensitivity of CVR mapping by applying a fast
imaging technology, multiband MRI sequence.
CVR mapping inherently has low sensitivity as it relies on small
BOLD signal changes in response to CO2 inhalation, similar to fMRI.
Recently, it has been shown that brain fMRI signal can be more robustly
measured when using a fast imaging technology called multiband
acquisition, and the technology has received wide attention. We
hypothesize that multiband acquisition can also improve the sensitivity of
vi
CVR data, by collecting images at a higher temporal resolution. In Aim 2,
we examined the benefit of multiband acquisition in CVR mapping by
comparing CVR data collected with multiband factor of 2 (imaging two times
faster) and 3 (imaging three times faster) to those with regular acquisition.
Aim 3: Investigate modeling and nonlinearity issues in CVR data
analysis.
The previous two aims examined data acquisition strategies in CVR
mapping. The present aim focuses on issues related to data analysis. A
linear relationship is usually assumed between EtCO2 and the BOLD signal,
making linear model the most widely used model in CVR analysis. However,
recent reports have suggested a nonlinear relationship between BOLD
signal and arterial CO2 concentration. In Aim 3, we proposed an improved
modeling scheme that incorporates possible nonlinearity while preserving
the linear effect, through which we investigated the extent of nonlinear effect
in CVR data and its dependence on age. | |