Session Item

MRI is useful to estimate respiratory motion in radiotherapy. Conventionally, motion estimation is done using either fast 2D imaging of a focal lesion or (retrospective) respiratory-correlated 4D acquisitions of the wider anatomy. Ideally, fast 2D imaging of the target for MLC tracking, and respiratory-correlated 4D imaging for dose accumulation could be performed at the same time. We propose a novel hybrid 2D/4D-MRI method to simultaneously facilitate real-time and 4D radiotherapy applications.

All experiments were performed on the 1.5 T Unity MR-linac (Elekta AB, Stockholm, SE) in research mode. The Quasar MRI4D phantom (ModusQA, London, CA) insert was moved using either Lujan motion (cos⁴, A=20 mm, T=4 s) or patient-derived respiratory motion (A=11 mm, T=3 s, 0.6 mm/min drift). It contains a 3 cm spherical target intersected by a film dosimetry cassette.
Our hybrid 2D/4D-MRI method uses a simultaneous multi-slice (SMS) accelerated sequence, which acquires two coronal slices simultaneously and repeatedly cycles through slices over the image volume. Based on self-navigation, respiratory-correlated 4D-MRIs can be reconstructed.
In this study, we enabled real-time streaming of all SMS 2D cine-MRI slices during beam-on. The SMS slice that intersected the insert was used to determine the target position relative to a mid-position reference derived from the pre-beam 4D-MRI, and the MLC was adjusted accordingly (Fig1).
Two SMS sequences (3.15 Hz T₂w turbo spin echo (TSE) and 2.35 Hz T₂/T₁w balanced turbo field echo (bTFE)) were acquired covering a stack of 28 slices (2x2.5x4.5 mm³ voxel size). For comparison, a conventional (fixed slice location) sequence (4 Hz T₁w TFE, 2.5x2.5x10 mm³ voxel size) was used as input for MLC tracking. End-to-end system latency for MLC tracking (w/ and w/o motion prediction) was estimated using the portal imager.
A 15-beam 8x7.5 Gy lung SBRT plan was created. Gafchromic EBT3 films were used to measure delivered dose. A local gamma analysis (2%/2mm) quantified dose differences between static and tracking deliveries, and a dose area histogram was used to assess the target coverage.

The total latency w/ (w/o) predictor was -1 (501), -10 (398), and 0 (319) ms for SMS-TSE, SMS-bTFE and conventional imaging, respectively. The target was underdosed without tracking, and compared to the static reference we found differences of up to 2 Gy (Fig2). The gamma pass rate increased from 51.4% to 96.3-99.7% (Lujan motion) and 63.0% to 98.4-99.7% (patient-derived motion) when tracking was applied. The target coverage of the tracking experiments was well above the prescribed dose and in excellent agreement with the static delivery: the D98%, D50%, and D2% target coverage agreed within 26, 19, and 31 cGy, respectively.

We demonstrated a novel hybrid 2D/4D-MRI method for MLC tracking on the Unity MR-linac. Comparable performance to conventional tracking was found, while our workflow provides the ability to reconstruct 4D-MRIs for dose accumulation.