Session Item

IMPT delivery accuracy may be compromised by setup and range inaccuracies and anatomical changes. These are mitigated by robust planning with setup robustness (SR) and range robustness (RR) settings and offline re-planning. Generally, SR settings are fixed for a tumor site and for the treatment course, using a value that ensures target coverage for the vast majority of patients. Daily re-planning could ensure adequate target coverage and reduce the required SR settings, but drastically increases workload. In this study we propose the use of a pre-treatment established library of treatment plans, each generated for a different SR. It was hypothesized that this online adaptive approach based on daily selection of the patient plan with the best fitting SR could improve CTV and OAR doses compared to conventional treatment with fixed SR settings. The concept was evaluated for head and neck cancer using computer simulations.

Weekly repeat CTs (rCTs) from 15 patients treated with 70 Gy_RBE to the primary CTV (CTV₇₀₀₀) and 54.25 Gy_RBE to the nodal CTV (CTV₅₄₂₅) were used. Contours were propagated from the planning CT (pCT) to the rCTs and manually adjusted if needed. For each patient, a plan library was generated based on the pCT using automated treatment planning. Plans were robustly optimized with 0, 1, 2, 3 or 5 mm SR settings and 3% RR settings.

For each patient, 25 treatment courses of 35 fractions were simulated with realistic setup and range errors. In the plan library approach, in each fraction a plan was selected from the library by evaluating the nominal dose distribution recalculated on the rCT. Both for conventional treatments and the plan library approach, CTV doses were accumulated. For OARs, NTCPs were compared.

In the plan library strategy, the 0, 1, 2, 3 or 5 mm plan was selected in 6%, 31%, 30%, 7% and 25% of the fractions respectively. 91.2% of the simulated treatments complied with all target aims, compared to 69.6%, 82.1% and 94.7% in the 2, 3 and 5 mm SR treatments, respectively (Fig. 1).

While maintaining overall adequate coverage, the plan library resulted in a mean reduction of 3.8 ± 2.1 (SD) %-point and 3.7 ± 2.5%-point for the risk of grade ≥ II xerostomia and dysphagia compared to using 5 mm SR (Fig.2). Compared to 3 mm SR, for 6/15 patients, the risk of xerostomia and/or dysphagia ≥ grade II could be reduced by > 2%. For the four patients (12-15) without NTCP improvement, adherence to target constraints was improved instead. Compared to 2 mm SR, adherence to target constraints was improved for 9/15 patients at the cost of NTCP.

The proposed plan library approach outperformed conventional treatment with fixed SR settings by reducing NTCP for similar coverage, or enhancing coverage without offline re-planning. Clinical implementation is rather straightforward making the approach a more practical online adaptive strategy than daily re-planning.