Magnetic resonance imaging-guided linear accelerator arterial spin labelling reveals dynamics of highly perfused non-enhancing glioblastoma during radiotherapy.

Background and Purpose: Targeting tumour vasculature during radiotherapy is a direction of current investigation for radiotherapy strategies in glioblastoma. Arterial spin labelling (ASL), a perfusion imaging technique that uses arterial water as an endogenous tracer, on magnetic resonance imaging (MRI)-guided linear accelerators (MRI-linacs) could guide adaptation to changes in perfusion. However, since ASL is unavailable as a stock sequence on MRI-linac systems, our objective was to implement the first MRI-linac ASL sequence, characterize its performance, and measure tumour perfusion dynamics. Materials and Methods: Forty-seven glioblastoma patients were imaged using 3D pseudo-continuous ASL on a 1.5 T MRI-linac during treatment. ASL labeling efficiency was measured in two healthy volunteers and three patients on the MRI-linac and a 1.5 T MR-simulation scanner. ASL cerebral blood flow (CBF) values and repeatability were characterized. Regions of high tumour CBF were evaluated for overlap with the gross tumour volume (GTV) and temporal dynamics. Results: The labelling efficiency was lower for the MRI-linac compared to the MR-sim and literature consensus values (0.59 vs. 0.88 vs. 0.85). Corrected grey matter CBF was comparable between the MRI-linac and literature (38 ± 13 ml/100 g/min vs. 36.5 ± 8.2 ml/100 g/min, p = 0.41). The within-subject coefficient of variation was similar to literature values (16 % vs. 11 ± 5 %). Across patients, approximately half of the high-CBF region did not overlap the GTV (median 47 %). The high-CBF region tended to decrease in volume during radiotherapy, from - 24 % at week 3 (p = 0.016) to - 45 % (p = 0.044) by week 5 relative to week 1. Conclusion: MRI-linac ASL could allow targeting and adaptation for highly perfused tumour in future trials for glioblastoma.