Abstract Book

S1065

ESTRO 37

Conclusion Quantitative dosimetric analysis was not able to show significant correlations between delivered dose, LET and RICV. However, no contralateral (asymmetric) fields’ arrangement approach seems to have an impact in the presence RICV toxicity. Asymmetric LET distributions, with high LET values at the contralateral distal edge of the PTV may contribute to RICV. EP-1959 Does library sub-categorisation improve auto- outlining accuracy in breast radiotherapy planning? C. Welgemoed 1 , E. Spezi 2 , M. Gooding 3 , D. Peressutti 3 , E. Aboagye 4 , R. McLauchlan 5 , D. Gujral 1 1 Imperial College Healthcare NHS Trust, Department of Radiotherapy, London, United Kingdom 2 Cardiff University, School of Engineering, Cardiff, United Kingdom 3 Mirada Medical Ltd, Research Department, Oxford, United Kingdom 4 Imperial College London, Department of Surgery and Cancer, London, United Kingdom 5 Imperial College Healthcare NHS Trust, Department of Medical Physics, London, United Kingdom Purpose or Objective Accurate definition of clinical treatment volumes (CTVs) and organs at risk (OARs) is essential to optimise breast radiotherapy outcomes and minimise radiation side- effects. Commercial software products have been developed for auto-outlining and have been shown to save time. In atlas-based systems, outlining accuracy has been show to improve when using atlas sub- categorisation. The aim of this project was to determine the difference in outlining accuracy, using two main atlas sub-groups (libraries) in breast radiotherapy. Material and Methods 22 Computed tomography (CT) data sets were used to retrospectively outline CTVs and OARs for patients who received radiotherapy to the left breast and nodes. Standard outlining guidelines were used for the brachial plexus, breast CTV and nodal volumes. Creation of one generic library consisting of 22 atlases and 2 main sub- groups with 3 atlas sub-categories each was performed by means of commercial software (WorkflowBox 1.4 Mirada Medical Ltd., UK). Breast separations (BS) from 22 atlases were divided into three atlas sub-categories: 0-19 cm, 20-23 cm and 24-30 cm, comprising of 6, 10 and 6 atlases respectively. Breast size categories were defined using the distance between the medial to lateral tangential field borders (BS). The same 22 atlases were sub- categorised according to breast wedge inclines: 20°, 15° and 10°, consisting of 1, 11 and 10 atlases each. Auto- outline accuracy compared to 3 “gold standard” reference outlines was measured using the Jaccard conformity index (JCI). Results The JCI comparison between breast wedge incline and breast separation sub-categories showed no significant difference. Median volume, JCI values, and centre of mass for the wedge incline and breast separation categories for the heart, brachial plexus, left breast CTV, humeral head, internal mammary chain, lung and inter- pectoral, level 1, 2, 3 and 4 lymph node CTVs are demonstrated in figure 1 (a-b) and figure 2 (a-c) respectively. JCI values for the heart, breast CTV, humeral head and lung were above 0.8, and considered clinically acceptable. JCI values for level 1-4 lymph nodes were greater than 0.5 and may be clinically acceptable with minimal editing .

Conclusion Clinically acceptable volumes can be automatically generated for the heart, breast, humeral head and lung as assessed by JCI. Similar results between atlas sub- categories were found, suggesting that in this initial study, categorisation had little effect on the atlas-based outlining system. Qualitative slice by slice assessments are required to confirm that JCI values reflect clinical assessment. EP-1960 Implementation of a statistical ideal DVH for the evaluation and optimization of treatment plans S. Bresciani 1 , F. Cresto 1 , A. Di Dia 1 , A. Maggio 1 , A. Miranti 1 , E. Garibaldi 2 , P. Gabriele 2 , M. Stasi 1 1 Candiolo Cancer Institute - FPO- IRCCS, Medical Physics, Candiolo TO, Italy 2 Candiolo Cancer Institute - FPO- IRCCS, Radiotherapy, Candiolo TO, Italy Purpose or Objective The aim of this study is to improve the quality and reduce the variability of treatment plans by developing a statistical ideal dose volume histogram (SI-DVH). From SI- DVH, we created a custom-built metric aimed to evaluate radiotherapic plans quantitatively through the definition of a series of scores varying on OARs and connected to linearly dependent functions. The SI-DVH is defined as the average DVH generated using the minimal dose that any voxel outside the targets must receive given 100% target coverage on a set of historical CT and countered volumes for prostate patients. Material and Methods CT images and structures of 37 prostate patients, 20 treated with TrueBeam Stx and 17 with Tomotherapy unit, were analyzed through the software PlanIQ (Sun Nuclear Corporation, USA). PlanIQ quantitatively defines regions of a DVH that are impossible, and difficult to reach, based on an ideal dose falloff from the prescription dose at the target boundary. It does not require any knowledge about the delivery technique but the minimal dose to OARs is predominantly dictated by the geometric relationship between OARs and PTV, and by the prescribed dose. After having acquired all the DVHs generated by software, we created for each OAR an average SI-DVH with feasibility equal to 0.1% (impossible) and 10% (difficult). Then we created the custom-built metric, based on 21 linear functions, to evaluate plan quality by applying a linearly dependent scoring system based on the QUANTEC constraints and on the SI-DVH; we defined PQM (Plan Quality Metric), sum of all the scores on target and OARs, and APQM (Adjusted Plan Quality Metric), the maximum ideally achievable score according