Session Item

Measurements were performed on a 1.5T Unity MR-linac (Elekta AB, Stockholm, SWE) using a Quasar MRI4D phantom (ModusQA, London, ON) and a custom-made ramp (20 deg. inclination) to support angulation (fig. 1). Unangulated, the phantom translated in CC, while angulated it translated in CC and AP. The phantom consisted of a water-filled body oval with a moving gel-filled cylinder in the middle. The cylinder contained a 3cm spherical target intersected with a cassette holding Gafchromic EBT3 film. 

Two motion traces were used: a linear 1mm/min drift (artificial) and a prerecorded trace of a prostate patient (real) (fig. 1). Clinical-grade 7-beam IMRT prostate plans (2x12Gy, 2mm isotropic GTV-PTV margins, 10 minutes delivery time) were made using Monaco (Elekta AB) for both angulation setups. 

MLC-tracking during delivery was performed by acquiring continuous 3Dcine MRI frames (0.1Hz, balanced gradient-echo, FOV 250x452x99mm³, voxels size 2x2x2.2mm³), that were rigidly registered using Elastix on the first static reference frame within a predefined mask around the target (fig. 1). Shifted MLC apertures were calculated and sent to the MR-linac via a research interface. The tracking performance was quantified by measuring the latency and RMSE of the tracked vs. phantom traces.  

Film dosimetry experiments were performed in which deliveries with/without MLC-tracking were compared with a static delivery. This was done for the real/artificial motion traces and both angulation setups. 

The tracking latency was 15±5s, including 500ms registration time, resulting in an RMSE for the artificial trace of 0.35mm. Selected dose maps (fig. 2) show excellent agreement between the static and tracked deliveries using the real trace with angulation. For all experiments combined, the local 2%/2mm gamma pass rates (median[min-max]) were 99[97-100]% for the tracked and 66[39-91]% for the non-tracked cases. Especially the artificial 1mm/min drift trace resulted in low pass-rates without tracking.