Regression algorithm correcting for partial volume effects in arterial spin labeling MRI

Partial volume effects (PVE) are a consequence of limited spatial resolution in brain imaging. In arterial spin labeling (ASL) MRI, the problem is exacerbated by the nonlinear dependency of the ASL signal on magnetization contributions from each tissue within an imaged voxel. We have developed an algorithm that corrects for PVE in ASL imaging. The algorithm is based on a model that represents the voxel intensity as a weighted sum of pure tissue contribution, where the weighting coefficients are the tissue's fractional volume in the voxel. Using this algorithm, we were able to estimate cerebral blood flow (CBF) for gray matter (GM) and white matter (WM) independently. The average voxelwise ratio of GM to WM CBF was approximately 3.2, in good agreement with reports in the literature. As proof of concept, data from PVE-corrected method were compared with those from the conventional, PVE-uncorrected method. As hypothesized, the two yielded similar CBF values for voxels containing >95% GM and differed in proportion with the voxels' heterogeneity. More importantly, the GM CBF assessed with the PVE-corrected method was independent of the voxels' heterogeneity, implying that estimation of flow was unaffected by PVE. An example of application of this algorithm in motor-activation data is also given.