ESTRO 38 Abstract book

S1051 ESTRO 38

therapy (IMPT) plans were created in twenty proxy patients with locally advanced lung cancer to a physical dose of 70Gy in 35 fractions. Proxy patients were selected to represent varying anatomical locations of the primary tumour and nodal involvements (15/20 had nodal involvement). Contouring, treatment planning and organs- at-risk constraints followed RTOG 1308 trial. The following cardiac sub-structures were delineated- right and left: atriums, ventricles and coronary arteries, and sino-atrial node. Dose calculation and optimisation of IMPT plans were done using Monte-Carlo dose engine. Dose to the heart and sub-structures were compared. Risk estimates of grade 3+ cardiac toxicities were calculated based on normal tissue complication probability mod- els which incorporated dose metrics and patients’ risk-factor - pre- existing cardiac disease (CD). Wilcoxon signed-rank test was used to assess statistical significance of the difference. Results There was no statistically significant difference in target coverage between VMAT and IMPT. Overall IMPT delivered lower doses to the heart (mean heart dose (MHD), V5 and V30). In VMAT plans, there were statistically significant positive correlations between heart dose and thoracic vertebral level (MHD, V5 and V30; Pearson correlation co- efficient, r: 0.67, 0.79, 0.48, P < 0.05). Between VMAT vs IMPT, there was no statistically significant difference in the mean cardiac dose or its sub-structures when the tumour (primary and nodes) extended above T7 vertebrae (n = 4). When tumour extended to and below T7 vertebrae (n = 16) IMPT delivered lower cardiac doses (MHD, V5 and V30, and mean dose to all sub-structures, P < 0.001). Risk of G3+ cardiac toxicities when tumour extended to and below T7 vertebrae are presented in Table 1. Conclusion Proton therapy has the potential to reduce cardiac toxicities compared to photon therapy. This analysis suggests that patients with tumour extension to and below T7 vertebrae would benefit most from proton therapy over photon therapy. The absolute benefit is higher in patients with underlying cardiac disease. EP-1932 Development of a deep learning network using a pre-trained convolutional neural network M. Rooney 1 , J. Mitchell 2 , D.B. McLaren 2 , W.H. Nailon 1,3 1 NHS Lothian, Department of Oncology Physics- Edinburgh Cancer Centre, Edinburgh, United Kingdom ; 2 NHS Lothian, Department of Clinical Oncology- Edinburgh Cancer Centre, Edinburgh, United Kingdom ; 3 The University of Edinburgh, School of Engineering, Edinburgh, United Kingdom Purpose or Objective The use of machine and deep learning is rising in oncology. Discrimination of tissue types using texture analysis is a long standing technique. Texture features are often used to train machine learning models. Deep learning, a subfield of machine learning, overcomes the need to calculate features by allowing the machine to learn directly from the image. However, a large amount of labeled image data is required to train deep learning models, this is a difficulty in oncology. The aim of this

Purpose or Objective Recent years have seen an increasing interest in studying the vascular response from large fractional doses delivered in stereotactic body radiotherapy (SBRT). While the vascular effect is extensively discussed as a potential source for tumour cell kill, the observed vascular response after doses of 10-15 Gy appears to be highly dynamic. With an initial reduction in blood flow that persists for various lengths of time, the actual vascular effect could be highly dependent on the time point at which treatment fractions are delivered with respect to each other. Thus, rather than leading to an increased cell kill, an increased radioresistance could be expected during the limited SBRT treatment if the total fraction of acutely hypoxic cells is increased. This study investigated the impact of temporary vascular collapse on tumour control probability (TCP) in SBRT, taking into account the different radiosensitivity of chronically and acutely hypoxic cells. Material and Methods Three-dimensional tumours were simulated based on tumour vessel distributions assuming different fractions of collapsed vessels at every treatment fraction. Thus, the simulated tumours contained both chronically and acutely hypoxic regions before the start of the treatment, the chronically hypoxic subvolume having a size of 30-60% of the tumour diameter, and a hypoxic fraction<5mmHg of 30-50%. The rest of the tumours were in general well- oxygenated at the start of the treatment. A radiation- induced increase in the acutely hypoxic fraction was simulated following the first fraction. SBRT-fractionation schedules of 3, 5, and 8 fractions were considered, and cell survival was calculated with a modified linear- quadratic model taking into account different radiosensitivities of chronically and acutely hypoxic cells. The simulated treatments were evaluated by calculating the TCP. Results A complex interplay between the radiation-induced and chronic hypoxia was observed for different fractionation schedules. For an eight-fraction treatment, for example, delivering 60 Gy in total, the TCP for a tumour with no treatment-induced vascular collapse in the well- oxygenated region and a chronically hypoxic subvolume with a diameter corresponding to 30% of the tumour size, was 97%. Assuming a vascular collapse of 35% induced by the first fraction and persisting throughout the remainder of the treatment resulted in a TCP of only 2% while the TCP was 83% for an identical tumour except for a much larger chronically hypoxic subvolume with a diameter of 60% of the tumour size. Thus, radiation-induced increase in acute hypoxia had a worse impact on the outcome of the treatment than an increase diameter of the chronically hypoxic subvolume. Conclusion The timing of SBRT fractions leading to vascular damages and thus increased acute hypoxia could impact on the efficacy of the treatment, and should be considered together with the tumour oxygenation to avoid loss of TCP for SBRT treatments. EP-1931 Photon vs proton therapy for reduction of cardiac toxicities in locally advanced lung cancer S. Teoh 1 , F. Fiorini 1 , B. George 1 , K.A. Vallis 1 , F. Van den Heuvel 1 1 CRUK/MRC Oxford Institute for Radiation Oncology, Oncology, Oxford, United Kingdom Purpose or Objective Identify a sub-group of patients with locally advanced lung cancer who would benefit most from proton therapy compared to photon therapy for reduction of cardiac toxicities using the model-based approach. Material and Methods Dual-arc volumetric modulated arc photon therapy (VMAT) and robust-optimised inten- sity modulated proton