Identification of arterial input function in perfusion imaging with MR angiography-supported semi-automatic method

Abstract

This thesis aims to improve arterial input function (AIF) selection in DSC-MRI by using the information gathered through magnetic resonance angiography (MRA) and cluster analysis of the concentration time curve (CTC) parameters. MRA was utilized with a dual-purpose, identifying arterial locations during the parametric evaluation of CTCs in DSC-MRI, and avoiding shape distortions in AIF. The knowledge of arterial locations is essential to the research, as it guided the cluster analysis carried out with the CTC parameters of voxels located within and around the middle cerebral artery (MCA). Additionally, it enabled us to identify the voxels that meet the AIF criteria and those with distorted CTCs. The literature has developed the following criteria for selecting AIF: high peak height (PH), small full-width-at-half-maximum, (FWHM), early time-to-peak (TTP), and early arrival time (AT). However, it has been found that high PH and small FWHM may indicate a shape distortion due to partial volume e ect (PVE). PVE is a common problem in AIF identi cation, which emerges when a voxel contains both artery and brain tissue. To avoid PVE, we included in our cluster analysis a recently introduced parameter, the SS:AUC ratio, which indicates the ratio of the mean steady state (SS) value (post-bolus equilibrium) to the area under the curve (AUC) of the rst passage of contrast agent. We calculated the SS:AUC of VOF and used it as a reference in selecting AIF. By using this reference value, we managed to detect the CTCs that were not distorted by PVE. If the SS:AUC of AIF was far from the reference value, CBF was either under- or over-estimated by a maximum of 41.1 14.3 and 36.6 19.2%, respectively.|Keywords : DSC magnetic resonance imaging, cerebral blood ow, arterial input function, MR angiography, cluster analysis, partial volume e ect, middle cerebral artery.