Session Item

We previously demonstrated the technical feasibility of MRI-guided multi-leaf collimator (MLC) tracking on the 1.5 T Unity MR-linac (Elekta AB, Stockholm, SE) using our own in-house developed imaging and treatment adaptation pipeline. Recently, Elekta installed pre-clinical automatic gating functionality on one of our Unity systems with a comprehensive motion manager (CMM) that continuously monitors 3D tumor motion. In this study, we perform MLC tracking based on the motion vector provided by the CMM to demonstrate a pre-clinical MRI-guided MLC tracking workflow for the first time. Additionally, we used a novel plastic scintillator detector (PSD) array to dosimetrically validate the tracking performance.

All experiments were performed on a Unity MR-linac that was equipped with pre-clinical automatic gating functionality. The CMM uses interleaved coronal and sagittal cine MRI (166 ms per slice) to estimate the 3D target positions. Using dedicated client software, the obtained positions were used to track the target by continuously moving the MLC aperture along with the target. A predictor (linear ridge regression) was used to compensate for the total system latency induced by the CMM and MLC tracking.

The Quasar MRI4D phantom (ModusQA, CA) was used statically, or programmed with patient-derived cranial-caudal motion (A=11 mm peak-to-peak, T=3 s, drift=0.3 mm/min). The phantom contained a dosimetry insert with Ø3cm spherical target (GTV). A 3x18 Gy IMRT plan with 3mm GTV-to-PTV margins was created following our clinical lung SBRT template. A 3D-printed cassette prototype with 8 PSDs (Medscint, CA) was placed coronally in the phantom’s dosimetry insert. The PSDs provide accurate time-resolved measurements at 10 Hz with instant dose read-out. The PSDs were positioned in the center of the target and at the cranial and caudal edges.

The total system latency was 347(±8) ms. The predictor reduces this to a negligible latency of 3(±5) ms.

Comparing the dose measured during a static delivery to the planned dose gave a dose difference of only 1.2% in the center of the GTV. Motion without tracking resulted in differences of up to 58% compared to the static delivery. Time-resolved PSD analysis revealed that dosimetric differences arose after 3.5 min, which coincides with an increase of drift motion in the motion trace. Tracking mostly restored the static dose distribution, prevented underdosage of the PTV, and limited hotspots outside the PTV.  

We demonstrated that CMM for gating is compatible with MLC tracking, facilitating a pre-clinical MRI-guided MLC tracking workflow that effectively mitigates respiratory motion during an IMRT delivery. The novel PSD array provided accurate time-resolved dosimetric measurements at key areas in the dose distribution, making it a valuable addition to validate (online) adaptive treatments.