New criterion for automatic AIF selection in DSC perfusion MRI to exclude partial volume effects

Introduction Partial volume effects (PVEs), which arise from the relatively low spatial resolution used in dynamic susceptibility contrast (DSC) MRI, often hamper the selection of the arterial input function (AIF, the evolution of the concentration contrast agent in a brain-feeding artery). PVEs have been shown previously to lead to shape changes in the AIF (1-3). More importantly, PVEs can lead to concentration profiles that are narrower and have a higher maximum than the true shape. This occurs when the tissue and arterial components of the signal at the peak concentration cancel out due to having opposite phases. Automatic and manual AIF selection procedures select voxels for the AIF measurement based on empirical rules (e.g. small width, high maximum and steep rise). These rules, however, do not exclude all voxels exhibiting PVEs, thus potentially leading to an erroneous measurement of the AIF. Since PVEs are inherently non-linear with respect to the concentration of contrast agent (1,2), the second passage of the contrast agent or the post-bolus equilibrium value (the ‘steady state’) provides additional information that can be used to identify AIF measurements affected by PVEs. In this study, we propose a method that exploits this extra information to identify AIF voxels that display minimum shape effects of partial volume (even if the partial volume itself might be present and the total scaling might vary (3)). Theory From tracer kinetics theory, it is known that: