Robust estimation of quantitative perfusion from multi-phase pseudo-continuous arterial spin labeling

Multi-phase PCASL has been proposed as a means to achieve accurate perfusion quantification that is robust to imperfect shim in the labeling plane. However, there exists a bias in the estimation process that is a function of noise in the data. In this work, this bias is characterized and then addressed in animal and human data.

The proposed algorithm to overcome bias uses the initial biased voxel-wise estimate of phase tracking error to cluster regions with different off-resonance phase shifts, from which a high-SNR estimate of regional phase offset is derived. Simulations were used to predict the bias expected at typical SNR. Multi-phase PCASL in 3 rat strains (n = 21) at 9.4 T was considered, along with 20 human subjects previously imaged using ASL at 3 T. The algorithm was extended to include estimation of arterial blood flow velocity.

Based on simulations, a perfusion estimation bias of 6-8% was expected using 8-phase data at typical SNR. This bias was eliminated when a high-precision estimate of phase error was available. In the preclinical data, the bias-corrected measure of perfusion (107 ± 14 mL/100g/min) was lower than the standard analysis (116 ± 14 mL/100g/min), corresponding to a mean observed bias across strains of 8.0%. In the human data, bias correction resulted in a 15% decrease in the estimate of perfusion.

Using a retrospective algorithmic approach, it was possible to exploit common information found in multiple voxels within a whole region of the brain, offering superior SNR and thus overcoming the bias in perfusion quantification from multi-phase PCASL.