Session Item

To quantify potential benefits of 4D robust optimization on multiple 4DCT acquisitions combined with off-line treatment adaptation (ART) for pre-operative carbon ion therapy (CIRT) of borderline resectable pancreatic adenocarcinomas (phase II clinical trial - NCT03822936).

For 10 previously treated patients, a 4DCT was acquired at -15 (CT_Plan), -5 (RV₁), 0 (RV₂) and +6 (RV₃) days from RT start. Treatment plans were optimized to a dose prescription of 38.4 Gy(RBE), in 8 fractions, with a constraint of 38 Gy(RBE) to 1% of the Gastrointestinal organs at risk (GI-OARs) volume (D_1%) and a plan objective of D_5cc<36 Gy(RBE). The employed motion mitigation strategy involves immobilization with a solid thermoplastic mask, gated dose delivery, centered on maximum expiration (0Ex), combined with 5 rescans. Three adaptive strategies were tested: (A) robust optimization on CTPlan-0Ex accounting for 3 mm set-up and 3% range uncertainty, including the 30%-inspiration phase; (B) robust optimization with the addition of RV₁-0Ex as uncertainty scenario; (C) plan recalculation at each RV_i and re-optimization (RP_i) according to pre-defined thresholds on dose deviation from clinical goals. The cumulative variation of target coverage and GI-OARs doses was evaluated, for each pipeline, assuming RV₂ and RV₃ representative of the subsequent 4 fractions. The duodenum contour of all available 4DCT for each patient was registered on CT_Plan. The capacity of pre-RT acquisitions to predict duodenum position, along the RT course, was investigated by computing the intersection of the contours at CT_plan, RV₁, or the union of both, with respect to post-RT 4DCT and the CTV, coupled with increasing margin expansions.

No RC_i ever exceeded the near-to-maximum (D_1%) constraint to GI-OARs (A). The use of robust optimization alone (B) improved CTV D_98% on average of (3.0±2.2)%, but was still sub-optimal on a patient-specific basis. According to (C), half of the plans would be re-optimized to recover target coverage and/or minimize duodenum D_5cc, at least at one RV_i (Figure 1). A significant difference was observed between duodenum contours on CT_plan, RV₁, or their union when intersecting subsequent contours, suggesting that the duodenum position could not be predicted by any combination of pre-RT contours, neither with a margin expansion, without substantially compromising target coverage (Figure 2). 

Figure 1: % Variation of CTV D_98% (a) and duodenum D_5cc (b) for each patient, according to the three strategies (A blue – B red – C green), at planning and at each RV_i with respect to optimization goals.

The off-line ART approach (C), currently applied to the pre-operative treatment schedule, proved feasible and safe. A fast and efficient online ART protocol is anyhow fundamental to improve treatment quality in CIRT for pancreatic adenocarcinomas.