Validating Kinetic Perfusion Parameter Mapping Using a High-Fidelity Whole Brain Simulation

Motivation: While experimental phantoms are useful for validating perfusion modeling, they lack the complexity of organs. Goal(s): To construct an in-silico whole brain perfusion model and use it for assessing the estimation errors of a two-compartment perfusion model. Approach: A large vessel segmentation obtained in a human subject is combined simulated arteries, veins and capillaries and their flow. Two-compartment modeling is then performed on the corresponding concentration maps computed and compared with the known ground truth. Results: The data demonstrate an underestimation of flow (mean of -18.40%), likely due to the delay and dispersion effects related to the choice of a global AIF. Impact: We demonstrate a detailed in-silico perfusion model that allows the investigate the accuracy of perfusion modeling methods.