A Linear Mixed Perfusion Model for Tissue Partial Volume Correction of Perfusion Estimates in Dynamic Susceptibility Contrast MRI: Impact on Absolute Quantification, Repeatability, and Agreement With Pseudo-Continuous Arterial Spin Labeling

Purpose: The partial volume effect (PVE) is an important source of bias in brain perfusion measurements. The impact of tissue PVEs in perfusion measurements with dynamic susceptibility contrast MRI (DSC-MRI) has not yet been well established. The purpose of this study was to suggest a partial volume correction (PVC) approach for DSC-MRI and to study how PVC affects DSC-MRI perfusion results. Methods: A linear mixed perfusion model for DSC-MRI was derived and evaluated by way of simulations. Twenty healthy volunteers were scanned twice, including DSC-MRI, arterial spin labeling (ASL), and partial volume measurements. Two different algorithms for PVC were employed and assessed. Results: Simulations showed that the derived model had a tendency to overestimate perfusion values in voxels with high fractions of cerebrospinal fluid. PVC reduced the tissue volume dependence of DSC-MRI perfusion values from 44.4% to 4.2% in gray matter and from 55.3% to 14.2% in white matter. One PVC method significantly improved the voxel-wise repeatability, but PVC did not improve the spatial agreement between DSC-MRI and ASL perfusion maps. Conclusion: Significant PVEs were found for DSC-MRI perfusion estimates, and PVC successfully reduced those effects. The findings suggest that PVC might be an important consideration for DSC-MRI applications. (C) 2016 International Society for Magnetic Resonance in Medicine