Comparison of different algorithms for minimizing macro vessel signal in cerebral perfusion imaging

INTRODUCTION: Dynamic-Susceptibilty-Contrast Magnetic Resonance Imaging (DSC-MRI) has been established as a reliable method to investigate hemodynamic changes in cerebral perfusion. Compared to PET and SPECT, DSC-MRI often overestimates of the amount of cerebral blood flow (CBF) and cerebral blood volume (CBV), which is due to the high sensitivity of DSC-MRI to large vessels [1]. To improve the correlation between DSC-MRI data and PET, a method referred to as ELV has been proposed. It consists of a defined threshold that is applied to CBF or CBV images to eliminate large vessel (ELV) signal [2]. Another ELV approach is based on a cluster analysis of the perfusion parameters to separate vessels from other tissue [3]. However, the ELV method does not take partial volume effects (PVE) into account and perfusion parameter maps suffer from the large number of excluded pixel. Carroll et al. [4] proposed a method to eliminate large vessel signal and reduce partial volume effects in DSC-MRI data using independent component analysis (ICA). The ICA method has been shown to be robust with regard to differences in temporal resolution and signal-to-noise ratio of the DSC-MRI time-series [5]. In this work we set out to compare the hemodynamic parameters CBF, CBV and the mean transit time (MTT) for uncorrected perfusion data with ELV and ICA vessel artifact corrected data.