Sequence- and Analysis Optimization for Partial Volume Correction in Arterial Spin Labelling based Perfusion MRI.
MetadataShow full item record
- Institutt for fysikk 
Arterial spin labelling (ASL) is a magnetic resonance imaging (MRI) based method that can measure quantitatively cerebral blood flow (CBF) by magnetically labelling the arterial blood that perfuse the brain tissue. The brain consists predominantly of three different components: white matter (WM), grey matter (GM) and cerebrospinal fluid (CSF), which differ in longitudinal relaxation times: T1, and flow. Due to the low spatial resolution in ASL, a signal from a given voxel is likely to consist of a mixture of signals from the different components weighted by their volume fraction. For accurate flow measurements, this partial volume (PV) effect needs to be taken into account.Recently, a method for obtaining PV maps based on multi-exponential analysis of the T1 recovery curves in an inversion recovery Look-Looker (IR LL) acquisition was proposed. In the current study, a modification of the IR LL sequence is presented. By introducing slice selective inversion pulses, magnetization history propagation between individual slices will be eliminated and thus significant reduction in acquisition time is enabled compared to sequences with non-selective inversions. Short acquisition time is an important factor to enable partial volume correction (PVC) in ASL in clinical applications since available scan time is limited. In the first part of the study, optimization of sequence parameters was performed based on both phantom and brain imaging. The optimal parameter combination was found to be RF pulses with flip angle = 4 deg, separated with ∆TI = 400 ms and with spoiling gradient amplitude of 10 mT/m after each read-out. Correcting for $B_1$-field inhomogeneity was shown not to be required and number of inversion times was set to n = 12. This provided viable PV maps with full-brain coverage with a total acquisition time of about 5 min, a 30 % reduction compared to the original IR LL sequence. In the second part of the study, PVC based on the PV maps from the optimal IR LL sequence was performed on ASL images. Two PVC methods, one based on thresholding and one based on a regression algorithm, were tested and compared with the raw uncorrected ASL signals. The ASL signals from the threshold-based PVC were found to be dependent of the voxel-wise PV fractions and thus suboptimal for PVC. In contrast, the regression analysis was capable of separating the tissue specific contributions and thus the resulting tissue specific ASL signals were independent of PV fractions. With reduced acquisition time, the applicability of the IR LL sequence for PVC in clinical ASL protocols has increased and can thus contribute to improvements of CBF measurements as a diagnostic and prognostic tool.