A novel Bayesian approach to accounting for uncertainty in fMRI-derived estimates of cerebral oxygen metabolism fluctuations

摘要: Calibrated blood oxygenation level dependent (BOLD) imaging is a multimodal functional MRI technique designed to estimate changes in cerebral oxygen metabolism from measured flow and the BOLD signal. This addresses fundamental ambiguities associated with quantitative signal analysis; however, its dependence on biophysical modeling creates uncertainty resulting estimates. In this work, we developed Bayesian approach estimating response neural stimulus used it examine that arises calibrated estimation due presence of unmeasured model parameters. We applied our CMRO2 visual task using traditional hypercapnia calibration experiment as well metabolic both measurement baseline apparent R2' technique. Further, order effects spinal fluid (CSF) contamination R2', examined measuring parameter without CSF-nulling. found two techniques provided consistent estimates average, median R2'-based CO2 1.4%, R2'- hypercapnia-calibrated 27% 24%, respectively. However, these were sensitive different sources uncertainty. The R2'-calibrated was highly CSF parameters describing flow-volume coupling, capillary bed characteristics, iso-susceptibility saturation blood. relatively insensitive but assumed CO2.