Clinicians, basic scientists and MRI technologists interested in stroke and cerebrovascular imaging.Time of Flight MRA (TOF) is currently used for intracranial artery assessment in patients with cerebrovascular disease. However, dephasing from slow/ turbulent flow and motion-related artifacts may impact TOF image quality and accuracy(1,2,3). Recently, a Non-enhanced Hybrid Arterial Spin Labeling MRA (NoHASL) technique has been described for the extracranial carotid arteries(4). It combines pseudocontinuous and pulsed arterial spin labelling, an undersampled 3D radial trajectory and a balanced steady-state free precession readout to maximise arterial contrast, coverage and robustness to motion. Our purpose was to evaluate feasibility, image quality and diagnostic assessment of NoHASL for the intracranial arteries in patients, with comparison to TOF and contrast enhanced MRA (CE-MRA).

30 patients (20M, 10F, mean 65y, range 21-91y) referred for MRI for known/ suspected cerebral ischemia underwent intracranial TOF, prototype NoHASL and CE-MRA at 1.5T (Avanto, Siemens Healthcare GmbH) during a single visit. 3D TOF (MOTSA) parameters were: TR/TE 23/7 ms, FA 25°, true voxel size 0.7x0.7x1.4mm3, FOV 180 x 180 mm, 3 slabs, 40 slices per slab, slice partial Fourier 7/8, acceleration factor 2 (GRAPPA), TA 3:29 minutes. NoHASL parameters were: TR/TE 3000/2.0 ms, echo spacing 3.9ms, FA 90°, true voxel size 0.8x0.8x0.8mm3, FOV 210 mm, TA 4:54 minutes, slices per slab 256, 7680 radial views, pseudocontinuous labeling duration 2000ms, post label delay 300ms. CE-MRA was performed from aortic arch to the circle of Willis: TR/TE 2.8/1.2 ms, FA 25°, true voxel size 0.9x0.8x1.1 mm3, FOV 410 x 308 mm, 120 partitions, 6/8 slice and phase partial Fourier, acceleration factor 3 (GRAPPA), TA 24s, using gadoterate meglumine (Dotarem, Aspen Pharmacare), 0.2ml/kg at 2ml/s.

Anonymized NoHASL, TOF and CE-MRA images were independently interpreted by two neuroradiologists (R1 & R2) on a PACS workstation (Impax, Agfa). 21 segments were assessed per subject (petrous ICA and intradural vertebral arteries to A2, M2 and P2 segments). Image quality (IQ) was scored (1=non-diagnostic, 3=sufficient for diagnosis, 5=excellent). Degree of stenosis was scored as non-significant (0-49%) and hemodynamically significant 50-100%. The reference standard was a consensus read of all 3 sequences by both raters. Mean±SD IQ for each sequence was compared with the Wilcoxon signed-rank test. Sensitivity / specificity and inter-rater agreement (kappa statistics) for haemodynamically significant stenosis were calculated.

All patients successfully completed all sequences. Overall IQ scores were diagnostic for all sequences: NoHASL 3.32±0.86, TOF 3.48±0.68 and CE-MRA 3.44±0.78, with mean TOF and CE-MRA scores both higher than NoHASL (p<0.01). At the segmental level, NoHASL performed well for the intracranial ICA, basilar artery and M1 segments, but mean IQ scores were below 3 (not sufficient for diagnosis) for smaller caliber vessels (Fig 1), largely due to low SNR and bright CSF signal. Reasons for TOF IQ scores < 3 were flow saturation, Venetian blind artifact and low SNR. Low SNR and small caliber vessels were cited as reasons for poor IQ scores <3 for CE-MRA.

Of 630 total arterial segments, 14 haemodynamically significant stenoses were detected at the reference standard (Fig 2 & 3). 5/14 stenoses were attributed to one motion-degraded study, with low SNR recorded for NoHASL and CE-MRA. For R1, NoHASL sensitivity/specificity for stenosis detection was 71.4%/84.4%, TOF 71.4%/96.6% and CE-MRA 42.9%/97.9%. For R2, NoHASL sensitivity/specificity was 64.3%/95.6%, TOF 64.3%/98.9% and 57.1%/99.5% for CE-MRA. When combining both raters the sensitivity/specificity for NoHASL, TOF and CE-MRA was 67.9%/90%, 67.9%/97.7% and 50%/98.7% respectively. There was no statistically significant difference detected in accuracy between sequences. There was fair inter-rater agreement for all three sequences, with kappa values of 0.30 (NoHASL), 0.37 (TOF) and 0.24 (CE-MRA).

NoHASL MRA is feasible, providing diagnostic quality imaging of proximal intracranial vessels. However, image quality was lower than TOF, particularly for smaller caliber vessels, largely due to poor SNR. Although no differences in accuracy were observed between sequences, there was low prevalence of steno-occlusive disease in the subjects studied. Our study also highlights the challenges of intracranial artery stenosis assessment, with only fair inter-rater agreement for all 3 techniques. Further NoHASL parameter optimization and implementation at 3T for greater SNR may improve its performance.