Abstract
Cerebrovascular reactivity (CVR) imaging is used to assess the vasodilatory capacity of cerebral blood vessels. While blood flow (CVR(CBF) ), blood velocity (CVR(v) ), and preferably blood volume changes (CVR(CBV) ) are used to represent physiological CVR, quantifying these measures is fraught with acquisition challenges in humans. Consequently, blood oxygenation level-dependent (BOLD)-MRI CVR (CVR(BOLD) ) is the most widely used MRI-based CVR method, even though it arguably provides the most indirect estimation of CVR. In this paper, we sought to holistically address the quantitative capacity and shortcomings ofCVR(BOLD) . To do so, we developed aCVR(BOLD) simulation framework and, together with data from theCVR(BOLD) literature, addressed whether and to what extentCVR(BOLD) accurately reflects CVR, and with which parametersCVR(BOLD) varies most. In short, we show the following:CVR(BOLD) does not necessarily correspond to physiological measures of CVR and depends on physiological (e.g., hematocrit) and acquisition (e.g., field strength) parameters;CVR(BOLD) is dependent on the stimulus protocol (e.g., breath-holding vs. controlled hypercapnia) chosen to elicit a vasoactive response; resting-stateCVR(BOLD) does not necessarily reflect breath-holdCVR(BOLD) , likely due to confounding neuronal activity; in stenotic disease and steal physiology,CVR(BOLD) results from a combination of factors which do not necessarily reflect the underlying CVR. We are confident that this work will provide researchers and clinicians with invaluable insights and advance the field of cerebrovascular imaging by enabling more accurate quantification of CVR in both health and disease.