Cyclotron and linear accelerator generated scanning proton beams for lung cancer SBRT: Interplay effects and mitigations.

Background: Pencil beam scanning (PBS) proton therapy for moving targets is known to be impacted by interplay effects between the scanning beam and organ motion. While respiratory motion in the thoracic region is the major cause for organ motion, interplay effects depend on the delivery characteristics of proton accelerators.

Purpose: To evaluate the impact of different types of PBS proton accelerators and spot sizes on interplay effects, mitigations, and plan quality for Stereotactic Body Radiation Therapy (SBRT) treatment of non-small cell lung cancer (NSCLC).

Methods: Twenty NSCLC patients treated with photon SBRT were selected to represent varying tumor volumes and respiratory motion amplitudes (median: 0.6 cm with abdominal compression) for this retrospective study. For each patient, plans were created using: (1) cyclotron-generated proton beams (CPB) with spot sizes of σ = 2.7-7.0 mm; (2) linear accelerator proton beams (LPB) (σ = 2.9-5.5 mm); and (3) linear accelerator proton minibeams (LPMB) (σ = 0.9-3.9 mm). The energy switching time is one second for CPB, and 0.005 s for LPMB and LPB. Plans were robustly optimized on the gross tumor volume (GTV) using each individual phase of four-dimensional computed tomography (4DCT) scans. Initially, single-field optimization (SFO) plans were evaluated; if the plan quality did not meet the dosimetric requirement, multi-field optimization (MFO) was used. MFO plans were created for all patients for comparisons. For each patient, all plans were normalized to have the same dose received by 99% of the GTV. Interplay effects were evaluated by computing the dose on 10 breathing phases, based on the spot distribution. Volumetric repainting (VR) was performed 2-6 times for each plan. We compared volume receiving 100% of the prescribed dose (V) of the GTV, and normal lung V.

Results: Twelve of 20 plans can be optimized sufficiently with SFO. SFO plans were less sensitive to the interplay effect compared to MFO plans in terms of target coverage for both LPB and LPMB. The following comparisons showed results utilizing the MFO technique. In the interplay evaluation without repainting, the mean V of the GTV were 99.42 ± 0.6%, 97.52 ± 3.9%, and 94.49 ± 7.3% for CPB, LPB, and LPMB plans, respectively. Following VR (2 × for CPB; 3 × for LPB; 5 × for LPMB), V of the GTV were improved (on average) by 0.13%, 1.84%, and 4.63%, respectively, achieving the acceptance criteria of V > 95%. Because of fast energy switch in linear accelerator proton machines, the delivery time for VR plans was the lowest for LPB plans, while delivery time for LPMB was on average 1 min longer than CPB plans. The advantage of small spot machines was better sparing in normal lung V, even when VR was applied.

Conclusion: In the absence of repainting, proton machines with large spot sizes generated more robust plans against interplay effects. The number of VR increased with decreasing spot sizes to achieve the acceptance criteria. VR improved the plan robustness against interplay effects for modalities with small spot sizes and fast energy changes, preserving the low dose sparing aspect of the LPMB, even when motion is included.