ESTRO 38 Abstract book

S996 ESTRO 38

This study aimed to develop a novel KBP model to reduce lung cancer plan variability, by predicting minimum achievable (or ideal) lung volume-dose measures V 5 /V 20 (volume of lungs receiving 5Gy/20Gy dose) and mean lung dose (MLD)) based on previously treated patients using our VMAT technique. The effect of the KBP model on treatment plan complexity and treatment delivery was assessed. Material and Methods 36 previously treated lung cancer patients were randomly selected from the Eclipse database and their dosimetric data and normal and target structures analysed. The three lung metrics, V 5 , V 20 and MLD, were correlated against residual lung volume (RLV) defined having iteratively searched for a lung volume construct that provided maximal correlation with all three metrics. It was found to be equal to the total lung volume excluding an isotropic 5 cm expansion of the PTV. A straight line was manually fitted to the correlation plots that demonstrated the lowest achieved dose metric values as a function of RLV. This was hence referred to as the lower-bound model. The model was tested by re-planning a further 39 patients, using the model predicted values as ideal constraints to replace protocol values. Treatment plan complexity metrics (MU/Gy, islands <1cc, small aperture scores) for the KBP plans were extracted. The effect of the models application on treatment deliverability was assessed by measurements. Treatment beams were measured with EPID panel and processed in portal dose image prediction software within Eclipse. Results A significant reduction in mean (maximum) V 5 , V 20 and MLD of 6.6% (19.8%), 1.1% (7.8%), and 0.7Gy (2.3Gy) respectively was achieved whilst maintaining optimal target coverage. The application of the model resulted in a 5.1% reduction in the number plans failing clinical constraints. A single observer study showed considerable reduction of variability in treatment plans. Plan complexity was observed to increase for KBP plans compared to the original plans, resulting in a small increase in measured delivery errors. Two of the 78 arcs failed the local optimal gamma pass criteria (3%/2mm≥98%) however they passed the hard criteria (3%/2mm≥95) and were therefore considered clinically acceptable.

When using VMAT optimization based on the gEUD, the following differences emerge: the programmed number of MU increases by over 40%; on average, the side of the equivalent square field decreases by more than 13%; the absorbed dose at the middle and lower levels of the rectum and bladder are considerably decreased. Conclusion The use of gEUD parameters produce an increase in the mean times for the optimization of the treatment plan, but a considerable OARs sparing. Despite the risk of complications is predominantly determined by the high doses, it is necessary not to neglect the control over the low- and mid-dose range. The optimization with cost functions based on the gEUD concept is straightforward and numerically expedient. In fact, with the only parameter gEUD you can control both high and low dose values; however, the need remains to better determine the parameter “a” for these OARs. References: Report of AAPM, The Use and QA of Biologically Related Models for Treatment Planning , (2012) TG 166 of the Therapy Physics Committee EP-1836 Validation of a novel knowledge-based planning (KBP) model for lung cancer treatments with VMAT N. Tambe 1,2 , C. Moore 1 , I.M. Pries 2 , C. Cawthorne 3 , A.W. Beavis 1,2,4 1 Hull and East Yorkshire Hospitals NHS Trust, Radiation Physics, Cottingham, United Kingdom ; 2 University of Hull, Faculty of Health Sciences, Hull, United Kingdom ; 3 KU Leuven, Nuclear Medicine and Molecular Imaging- Department of Imaging and Pathology- Biomedical Sciences Group, Leuven, Belgium ; 4 Sheffield-Hallam University, Faculty of Health and Well Being, Sheffield, United Kingdom Purpose or Objective Radiotherapy plan design can vary widely and is dependent on the experience of the treatment planning staff. Plans meeting planning objectives may still be suboptimal where there is scope to reduce OAR doses without compromising target coverage and deliverability.

Conclusion Our study showed that our novel in-house KBP models allowed a reduction in heterogeneity in treatment plan quality, with a concurrent decrease in dose to lung volume for VMAT-treated patients. Treatment delivery verifications demonstrated the KBP model plans were deliverable. The model has been implemented clinically using Eclipse scripting. EP-1837 A new hybrid approach to allow robust Monte Carlo-based multi-field optimization in proton therapy F. Tommasino 1 , L. Widesott 2 , F. Fracchiolla 2 , S. Lorentini 2 , R. Righetto 2 , C. Algranati 2 , E. Scifoni 3 , F. Dionisi 2 , D. Scartoni 2 , D. Amelio 2 , M. Cianchetti 2 , M. Schwarz 2 , M. Amichetti 2 , P. Farace 2 1 University of Trento, Physics, Trento, Italy ; 2 Azienda Provinciale per i Servizi Sanitari APSS, Protontherapy Department, Trento, Italy ; 3 Istituto Nazionale di Fisica