Estimation of Cerebral Perfusion Parameters Using Bayesian Probability Theory
Shimony J, Lee J, Derdeyn C, Powers W, Videen T, Markham J, Snyder A, Bretthorst G
Washington University Medical School
Dynamic susceptibility contrast MR methods have several drawbacks that limit its accuracy. These drawbacks include using a single arterial input function (AIF) for the entire brain and the need to perform a numerical deconvolution on the logarithm of noisy data. We implemented a tissue perfusion model that has several advantages over standard methods, including the estimation of a separate AIF for each pixel and estimation of additional perfusion parameters of clinical interest. The model parameters were estimated using Bayesian probability theory in a group of patients with hemodynamic impairment and compare favorably with PET and standard DSC perfusion estimates.