ESTRO 36 Abstract Book

S258 ESTRO 36 _______________________________________________________________________________________________

OC-0487 Pre-treatment characteristics can predict anatomical changes occurring during RT in lung cancer. L. Hoffmann 1 , A. Khalil 2 , M. Knap 2 , M. Alber 3 , D. Møller 1 1 Aarhus University Hospital, Department of Medical physics, Aarhus, Denmark 2 Aarhus University Hospital, Department of Oncology, Aarhus, Denmark 3 Heidelberg University Hospital, Department of Oncology, Heidelberg, Germany Purpose or Objective Anatomical changes such as the resolving atelectasis seen in Fig 1. prompt adaptive radiotherapy (ART) for a large number of lung cancer patients in order to avoid target under dosage. ART may re-establish the original dose distribution on the cost of additional work load. We investigated the correlation between patient characteristics before RT and anatomical changes during RT in order to identify the patients eligible for ART.

(CBCT)], which is registered to the planning CT (pCT) using a rigid body alignment. In the presence of non-rigid anatomical changes, it is not obvious which isocenter shift is the best with respect to target coverage and normal tissue sparing. We evaluate an alternative approach, where the dose is recalculated on daily scatter-corrected CBCT (scCBCT) images and the isocenter shift is determined using an interactive multicriterial optimization of DVH objectives. Material and Methods To enable dose calculations, the CBCT projections were scatter corrected using forward projections of the virtual CT (deformable image registration of pCT on CBCT) as a prior (Park et al. 2015, Med Phys). PTV and OAR structures were transferred from the pCT to the scCBCTs and corrected by an experienced clinician. In MIRA, a research planning system interpolation between pre-calculated sample dose distributions for a set of 13 isocenter positions allows navigating continuously on the set of Pareto-optimal isocenters. DVH parameters can be manipulated interactively. The resulting isodose lines and integral DVHs of this trade-off are displayed in real time, allowing the user to repeatedly manipulate the parameters until the clinically optimal solution is found. For the resulting isocenter shift, a final dose calculation is performed. The approach is evaluated for an exemplary head and neck (H&N) patient case. The prescribed dose was 54Gy in 30 fractions with 2 integrated boosts of 60Gy and 66Gy, respectively. For 5 scCBCTs the optimized dose distribution was compared to the ones of the clinically applied shifts. To evaluate the accuracy of the underlying dose interpolation, 100 random isocenter shifts for each of the scCBCTs were interpolated and compared to an MC calculation using a 2%/2mm gamma criterion. Results Dose interpolation accuracy was high [median gamma pass rate: 99.0% (range 96.6-100.0%)]. The spinal cord D 2% was comparable for both approaches (mean change -0.2Gy, range -1.7 to 0.2Gy). The mean dose of the parotid glands could be improved for 2 out of 5 fractions (one of them is displayed in Fig. 1), for the other 3 it could be preserved (mean change -1.0Gy, range -2.2Gy to +0.4Gy). Target coverage was preserved. The mean Euclidean distance between the clinical and the optimized isocenter was 1.8mm (range 0.8-3.2mm).

Material and Methods A decision support protocol for ART was used for treatment of 165 lung cancer patients. The patient setup on the primary tumour (T) was based on daily pre- treatment cone-beam CTs. Deviations in T >2mm, lymph nodes (N) >5mm or changes in atelectasis (A) or pleural effusion (PE) triggered replanning. The daily CBCTs were retrospectively reviewed to score changes above trigger limit in T or N position/shape, changes in A or PE, as well as T or N shrinkage >1cm. The findings were correlated to pre-treatment patient characteristics as histology, T and N volume, location and number, A or PE, and T or N adjacent to or surrounding the bronchi. Fisher’s exact test was used for comparison. p<0.05 was considered significant. Results Fifty three patients (32%) were adapted due to changes in A (8%), T (6%), N (15%) or T+N (3%), and 7% had more than one replan. Atelectasis was seen at planningCT in 50 patients (30%) while in 12 patients (7%), it appeared during RT. Presence of A before or during RT was not significantly correlated with replanning. However, the changes in A during RT significantly increased the probability of replanning. (p=0.03), see Fig.2. Additionally, A within 5mm of T or N was significant (p=0.01). Only 11 patients (6%) had changes in PE, but only in one patient was replanning indicated. Patients with T0 or N0 had a significant low risk of replanning (p=0.01, p=0.03) while patients with two or more N had a high rate of replanning (ns). Nodes in stations 1,2,3 or 4,7 or 10,11,12 had significantly higher rate of replanning as compared to patients with nodes in station 5,6 and 8,9. Node-volume >30cm 3 had a significantly higher rate of replanning (p=0.02). No correlation was found for T location, T size, T or N adjacent to bronchi or for T or N shrinkage. Histology was not significant for replanning. The imaging rate may be decreased for patients with T0 and no A, as none of these were adapted. For patients with N0 and no A, only 11%, were replanned. On the contrary, 60% of the patients with A and N volume>30cm 3 were replanned.

Figure 1: Comparison between a clinical (dashed) and an optimized shift (solid). The mean dose to the left parotid gland improved from 30.6 to 26.1Gy. Conclusion Compared to a rigid bony alignment, the novel, interactive, DVH based positioning offers increased control over OAR dose and PTV coverage. For a first H&N case, in some fractions the dose to the parotid gland was improved. Acknowledgements: DFG-MAP and BMBF-SPARTA