Arterial spin labelling (ASL) is a powerful method for non-invasively measuring blood flow in organs such as the brain. It works by labelling blood as it passes a tagging plane and then imaging the transit of the labelled water into the parenchyma of the organ of interest.

High quality ASL images are dependent upon a uniform magnetic field, not only in the imaging plane, but also in the tagging plane. Off-resonance effects arising from poor shims and patient geometry lead to differing labelling efficiency in each vessel in the tagging plane which ultimately leads to poor quality perfusion maps. This is a particular problem in pre-clinical imaging studies for two reasons: (1) rodent heads have air spaces very close to the tagging and imaging planes leading to susceptibility artefacts (throat, oesophagus, mouth and nasal cavities); and (2) the high field strengths used (typically ≥7T) make achieving good quality shims harder.

One solution to this is to implement a multiphase ASL sequence¹, where instead of collecting only two phases (here 180° and 0°), images are collected at many phases. By doing so, each voxel can be fitted individually to minimise off-resonance artefacts.

Rats were anaesthetised with isoflurane and imaged in a 9.4T MRI spectrometer (Agilent) using a 72mm volume transmit coil and a 4-channel surface receive array (Rapid Biomedical). The tagging plane was 6.2mm thick and positioned in the neck. Imaging readout was multislice single-shot spin echo EPI with FOV=32x32mm (64x64 matrix, thickness=1mm), TR=4s, TE=12.4ms, labelling bolus duration=1.4s (a train of 600µs 40° Hanning-shaped pulses, starting every 1.2ms; 50% duty cycle). Multiphase pseudo-continuous ASL (pCASL) was implemented using 8 phase angles, each 45 degrees apart and spaced from 0 to 315°. To allow modelling of bolus arrival time, 12 post-label delays from 50ms to 2s were acquired, each with 8 phase angles. Total acquisition time was 12m57s, reducible to 73s if bolus arrival time maps were not required. The imaging time for only tag and control images with a single post-label delay was 25s.

Tagging plane localisation was determined using time of flight angiography to visualise vessels and anatomical MRI to visualise the brain. Angiography: GE3D readout with TR=30ms, FA=30°, FOV=40x40x60mm (128x128x192) with an axial excitation slab, acquisition time 12m17s. Anatomical MRI: FSEMS readout with TR=1s, TE=10ms, FOV=40x60mm (128x192) single 2mm slice at brain midline, acquisition time 2m8s. Data from eight phases were fitted to a modified Fermi function shown below with α=66 and β=21 (chosen by fitting the function to experimental datasets) using a modified version of BASIL² before processing with oxford_asl³. Reference scans acquired with no tagging were used for absolute quantitation of CBF.

$$f(x) = -2 \left[\frac{1}{1+e^{(|x|-\alpha)/\beta}}\right]+1$$

The tagging plane location across the vessels in the neck and imaging slice locations across the brain are shown in Figure 1. Example perfusion maps obtained by fitting eight-phase data before producing the perfusion maps are shown in Figure 2. To compare to the multiphase data and to simulate off-resonance effects, four pairs of images, each taken 180° apart but taken at 45° offsets were processed as if they were single tag and control pairs (Figure 3).

Off-resonance effects in pre-clinical ASL can lead to poor estimation of perfusion, often systematically affecting entire vessel territories. This is problematic for interpretation of images in many contexts e.g. stroke imaging where entire territories are already hypoperfused or in tumours where compromised vasculature may affect perfusion within the tumour as well as downstream.

The use of multiphase imaging adds only 48 seconds to image acquisition yet allows each voxel’s specific phase shift to be fitted, considerably improving the quality of achievable perfusion maps. Acquisition of each phase at multiple post-label delays also allows fitting of bolus arrival time

Rodent ASL quality is readily improved at high field strengths by implementing multiphase ASL and fitting each voxel separately.