Y. Zhang¹, W. Tan², and M. K. H. Chan³; ¹Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, ²Department of Oncology, Shenzhen Hospital of Southern Medical University Shenzhen, Shenzhen, Shenzhen, China, ³Department of Radiation Oncology, University Nebraska Medical Center, Omaha, NE

Purpose/Objective(s): 3mm robustness set up uncertainty is generally used in clinic for head and neck cancer robust optimization (conventional robust optimization, cRO3mm). The 3mm considered the inter-fraction anatomical variability since the setup error can be controlled in 1mm in practice. However, anatomical variations are not rigid and uniform across all directions. The cRO3mm can lead to overdose on the organs at risk. We instead include predicted anatomical changes and reduce the setup margin to 1mm as geometric uncertainty for robust optimization (aRO1mm). The dosimetric distribution of aRO1mm is compared with cRO3mm. Materials/

Methods: 5 nasopharynx radiotherapy patients were included in this retrospective study, each with validation CT scans (vCT) from fraction 5 to 35 with 5 as interval. The predictive anatomical model, which is built to capture the average deformation of population at each time interval, are used to provide the predicted CT of fraction 5, 15 and 25.Three plans with different optimization scenarios are made in this study. 1)Three predictive images with 1mm as setup shift and 3% range uncertainty, and 2) robust 3mm setup shift and 3% range uncertainty, and 3) robust 1mm setup shift and 3% range uncertainty are used for aRO1mm, cRO3mm and cRO1mm, respectively. Accumulated nominal and voxel-wise minimum (voxmin) dose of vCTs under 1mm setup and 3% range uncertainty are calculated for analysis.
Results: The accumulated V94_voxmin of CTVs in cRO3mm are all above 98% (clinical goal). The mean value of ratio (accumulated V94_voxmin/nominal V94_voxmin) in Ao1mm is the highest among three strategy for high-risk CTV, but has a case whose accumulated V94_voxmin is 96.44%, slightly lower than clinical goal. While V94_voxmin of high-risk CTV for this case in cRO1mm is 94.05%. For low-risk CTV, 1/5 (97.68%) case is lower than 98% in aRO1mm, while 3/5 cases is lower than 98% in cRO3mm. The comparison of the nominal mean doses (Dmean) of OARs between Ao1mm and cRO3mm are shown in table I. The benefit of reduced dose on parotid and submandibular glands reflected on the estimated normal tissue complication probability (NTCP) [1] of moderate to severe dry month is 4%, and the benefit of reduced dose on oral cavity and PCMs reflected on the NTCP of swallowing problems (only soft food) is 1.5%.
Conclusion: We demonstrated that aRO1mm can substantially reduce the dose on the OARs compared to the conventional way with good target coverage. It provides a better way of controlling the uncertainty. reference: J. Langendijk, F Hoebers, M de Jong, P Doornaert, et al. National Protocol for Model-based Selection for Proton Therapy in Head and Neck Cancer. Int J Part Ther 2021. Abstract 2234 – Table 1: Comparison between Co3mm and Ao1mm on average Dmean of OARs over 5 patients. Unit is Gy.

Plan	ipsilateral parotids	contralateral parotids	submandibular glands	oral cavity	pharyngeal constrictor muscles (PCM) inferior	PCM middle	PCM superior
cRO3mm	30.47	25.09	46.04	17.85	22.01	33.82		54.85
aRO1mm	26.53	21.37	40.73	16.44	17.45	27.76		53.70