Abstract book - ESTRO meets Asia

S42 ESTRO meets Asia 2018

1 Liverpool and Macarthur Hospitals, Cancer Therapy Centres, Sydney, Australia 2 Ingham Institute, Applied Medical Research, Sydney, Australia 3 Australian e-Health Research Centre- CSIRO, Royal Brisbane Hospital, Brisbane, Australia 4 Calvary Mater Hospital, Radiation Oncology, Newcastle, Australia 5 Wollongong Hospital, Illawara Cancer Care Centre, Wollongong, Australia 6 Princess Alexandra Hospital, Radiation Oncology, Brisbane, Australia Purpose or Objective Inter-observer variation in volume delineation is a common challenge contributing to the global uncertainty in radiotherapy. The impact of this uncertainty in clinical trials is substantial, reducing the power of a study and limiting translational outcomes to future patients. Manual review of planning volumes is arduous, costly and resource intensive. For that reason, we propose a scalable cloud- based method known as, ‘The Radiotherapy Atlas Contouring (TRAC) tool,’ to improve accuracy and consistency in radiotherapy contouring in a clinical trial setting. Material and Methods For this study, retrospective data (n=93) from the prostate SBRT trial PROMETHEUS (ACTRN12615000223538) was utilised and it is one of the four trials explored in this project. 10 patients MRI data was used where five observers contoured three anatomical structures (bladder, rectum, prostate) and a rectal stabilisation device (SpaceOAR or Rectafix) to be regarded as the ‘benchmark contours’ for multi-atlas generation. A random forest classifier was trained using a range of contour parameters (DSC, Components, Volume, Elongation, Perimeter, Roundness, SphericalRadius, EquivSphericalPerimeter, Flatness) to detect pass/fail on the clinically treated versus ground truth contours using tenfold cross validation. Accuracy was determined based on the number of times the model correctly identified pass/fail contours. Results Analysis time was approximately 2 seconds per contour, far quicker than manual review. The TRAC tool was tested on 93 individual trial patients, the number of cases for training varied from 38 to 80. For patients treated with the Rectafix, the TRAC tool demonstrated accuracy (k=6) results (mean±SD) of 90.0±0.6%, 57.4±1.2% and 84.8±0.4% for the bladder, rectum and prostate respectively. For patients treated with the SpaceOAR the TRAC tool demonstrated accuracy results of 76.8±0.9%, 72.9± 0.1% and 67.6±0.1% for the bladder, rectum, and prostate respectively. Conclusion This proposed automated novel approach will deliver efficiency gains and a reproducible method for validating contouring consistency along with enhancing translation of trial results to the clinic to improve overall treatment outcomes. Results reported here are for limited training data and could be improved upon with more data and refinement. Additional testing is essential for other treatment sites and imaging modalities. OC-111 Management of radiation dermatitis using low-level laser therapy in NPC patients: A pilot study S.W.Y. Lee 1 , Y.W. Chan 2 , V.W.C. Wu 1 , C.L. Chan 2 , V.C.W. Tam 3 , D.L.W. Kwong 4 1 The Hong Kong Polytechnic University, Department of Health Technology and Informatics, Kowloon, Hong Kong SAR China 2 The University of Hong Kong, Department of Surgery,

Hong Kong, Hong Kong SAR China 3 Prince of Wales Hospital, Department of Clinical Oncology, New Territories, Hong Kong SAR China 4 The University of Hong Kong, Department of Clinical Oncology, Hong Kong, Hong Kong SAR China Purpose or Objective To evaluate the effectiveness of using low-level laser therapy (LLLT) in alleviating radiation-dermatitis (RD) in nasopharyngeal cancer (NPC) patients receiving concurrent chemo-radiotherapy (CRT). Material and Methods 10 NPC patients who received radical CRT (2Gy/fr, 5 fr/week to 70Gy) were recruited in this randomized triple- blinded, SHAM-controlled study. All patients acted as their own control. LLLT was given 30-60 minutes prior to radiotherapy (RT), twice a week throughout the course for 7 weeks. With the laser probe placed in the neck and chin of the designated treatment side, 3-4 shots (1-minute/shot) was given. The SHAM treatment was given on the opposite side, while both operator, patient and assessor were blinded to the side of treatment. Radiation skin reactions were graded weekly by a plastic surgeon using the RTOG acute radiation morbidity scale until one month after completion of RT.

Results Data were analysed for 8 patients. In 6 out of 8 cases, the severity of RD was minimal (RTOG G0) on the treatment side, and RTOG G1 erythema on the Sham treatment side by week 8. The surface area of RD in the neck region was significantly smaller on the treatment side in the remaining 3 cases. One case was found to have G3 reaction on both sides of neck at week 9. All patients did not report any adverse effects directly attributable to LLLT. Conclusion A significant reduction in the incidence of RD (100% SHAM vs 25% LLLT) suggested that LLLT has the potential to minimize the degree of skin reactions, thus warrants further investigation. OC-112 Comparison of two different atlas-based auto segmentation approaches for prostate cancer T. Aoyama 1 , H. Shimizu 1,2 , K. Yokoi 1 , H. Tachibana 1 , T. Kodaira 1 1 Aichi Cancer Center, Radiation Oncology, Nagoya, Japan 2 Gunma Perfectural College of Health Sciences, Graduate School of Radiological Technology, Maebashi, Japan Purpose or Objective Recently, atlas-based auto segmentation (ABS) software has become commercially available. Several reports have shown that ABS software saved time and decreased interobserver variation in contouring, but the default ABS approach was used. It is unclear if the results reflected the best achievable from the ABS software or if different