Improved partial volume correction for single inversion time arterial spin labeling data.

Arterial spin labeling has relatively low spatial resolution, which affects cerebral blood flow measurements by partial volume effect occurring at tissue interfaces, e.g., between gray matter, white matter, and cerebrospinal fluid. This can be an important source of cerebral blood flow quantification error. To correct for partial volume effect in arterial spin labeling, a linear regression method was recently proposed. Because this method assumes that tissue magnetization and cerebral blood flow are constant over an n(2) × 1 regression kernel, an inherent spatial blurring is introduced. In this study, a modified least trimmed squares algorithm is proposed for partial volume effect correction. It is demonstrated using simulations that the modified least trimmed square method can correct for partial volume effect and produce less blurring than the linear regression method. This is achieved without either acquiring additional datasets or increasing the computation burden. These capabilities were further demonstrated in vivo. The modified least trimmed square method should, therefore, play an important role in arterial spin labeling studies.