Session Item

To maximize healthy tissue sparing in the presence of intra-fractional motion, we previously developed MRI-guided MLC tracking for the 1.5T Unity MR-linac (Elekta AB, Stockholm, SE). Dosimetric analyses were performed using film dosimetry which lacks temporal and through-plane spatial resolution. To expand the dosimetric analysis of MLC tracking from 2D to time-resolved 3D, and to eliminate film-related dosimetric uncertainty, the Delta4+ MR phantom (ScandiDos AB, Uppsala, SE) is a viable alternative. However, the Delta4 cannot move and its electronics do not allow for real-time MR-imaging. In this study, we introduce a virtual HexaMotion platform for the Delta4 to quantify the dosimetric benefits of MRI-guided MLC tracking.

All experiments were performed on an MR-linac in research mode. For the Delta4 we used research software provided by ScandiDos enabling VMAT support and time-resolved data export/input.

Because the Delta4 cannot move, we initially used the Quasar MRI^4D phantom (ModusQA, London, ON) for image recording and programmed it with patient-derived respiratory CC motion (A=11mm, T=3s, drift=0.6mm/min). We pre-recorded the phantom motion using 2D-cine MRI at 4Hz, and simultaneously logged the phantom reported (ground-truth) positions. Next, to perform tracking with the Delta4, we streamed the recorded Quasar images to the tracking software to mimic a delivery to a moving target.

The Delta4 was positioned centrally or peripherally (10cm off-centre) in the bore. The phantom contains two orthogonal planes (sagittal/coronal) filled with diode arrays to measure dose every 25ms. The central 6x6cm² of diodes are spaced 5mm apart, with 10mm spacing elsewhere. Because the Delta4 records the received dose every 25ms, and the logged target position is known within each dose interval, we can deduce how far off-centre the dose was delivered. To artificially move the phantom with the target, we shifted the measured incremental dose by the same amount but in opposite direction as the target had moved w.r.t the iso-centre.  Dose shifts were carried out by re-gridding the data using cubic spline interpolation.

We created three VMAT treatment plans for lung SBRT with 3mm GTV-to-PTV margins: a central plan (8x7.5Gy) and two peripheral plans (3x18Gy and 1x34Gy). A 1%/1mm local Gamma-analysis quantified dose differences between a static delivery and tracking cases. 

The 1x34Gy dose maps (Fig 2) show the effect of a virtual HexaMotion platform. W/o tracking (no re-grid) differences with the static delivery are up to 10Gy, while this reduces to <1.5Gy with tracking. Following a single diode over time shows superimposed dose lines for static and tracking, while w/o tracking it receives only 60% of the dose.

Gamma pass-rates improved from 74.0% to 98.4% (8x7.5Gy), 34.7% to 98.1% (3x18Gy) and 17.8% to 99.7% (1x34Gy).

We created a virtual HexaMotion platform to quantify the time-resolved performance of MRI-guided MLC-tracking in 3D using a Delta4 on Unity.