ESTRO 38 Abstract book

S626 ESTRO 38

efficient clinical placement and reduce overall clinical placement time. In turn, this can increase student throughput and capacity. PO-1128 Clinical implementation of deformable image registration (DIR) C. Callie 1 , G. Dinsdale 1 , S. Deshpande 1 , M. Jameson 2 1 SWSLHD Radiation Oncology, Liverpool and Macarthur Cancer Therapy Centres, Liverpool BC, Australia ; 2 Ingham Institute for Applied Medical Research, Ingham Institute, Liverpool, Australia Purpose or Objective Deformable Image Registration (DIR) is increasingly used worldwide in adaptive radiotherapy scenarios, although local knowledge and experience within Australia is limited. DIR is a complex technique and clinical use of it brings many uncertainties, potential risks and workflow considerations. A recent publication by AAPM Task Group 132 (TG-132) addresses this and has outlined recommendations for mitigating potential risks associated with image registration. DIR involves significant collaboration between the multidisciplinary team, therefore governance around responsibilities and communication is essential. Training and education amongst all staff groups is also important. The purpose of this presentation is to discuss our departments’ experience in the clinical implementation of DIR for pre- planning head and neck PET fusions. This will cover workflow, quality assurance procedures, education and incorporation of TG-132 recommendations. Material and Methods MiM Maestro software V6.8.5 was used as the platform to perform DIR before the data was transferred to either Tomotherapy or Pinnacle for planning. Multidisciplinary stakeholder meetings were held between radiation oncologists, radiation therapists and medical physicists to determine the clinical workflow and governance of data management. An in-house quality assurance tool was developed to provide quantitative information on the results of the deformation in user defined regions of interest. The TG-132 report was reviewed, and processes were developed to incorporate the recommendations in to clinical practice. A comprehensive multidisciplinary competency package was developed to address training and education requirements. Radiation oncologists, radiation therapists and medical physicists were trained using our competency based system. The requirements to be deemed competent were tailored for each group. This involved: watching training videos, completing quizzes, performing practice cases and attending DIR registrations. Results A DIR request was developed and added to our existing booking process in Mosaiq Oncology Information System. Quality Checklist items (QCLs) were created in Mosaiq to manage workflow and facilitate communication between the multidisciplinary team (figure 1). Processes were established to combine the report generated from the in- house QA tool with a modified MiM report to meet TG-132 recommendations. DIR training cases were loaded into a training database. Radiation therapists performing DIR completed three cases of varying difficulty as part of their competency requirements. Radiation oncologists were required to attend three DIR registrations to understand the process before being able to freely request the technique for their patients. Medical physicists received training and performed initial cases under close supervision by lead physicist.

Conclusion DIR was successfully implemented for head and neck pre- planning PET and PETCT within our department. PO-1129 An analytical approach to aggregate patient workflows for system dynamics modelling of radiotherapy J. Lindberg 1 , P. Holmström 1 , S. Hallberg 2 , T. Björk- Eriksson 3 , C. Olsson 1 1 Clinical Sciences, Radiation Physics, Gothenburg, Sweden ; 2 Templog, Gothenburg, Sweden ; 3 Clinical Sciencies, Oncology, Gothenburg, Sweden Purpose or Objective Radiotherapy (RT) is one of the most technology-intense and complex disciplines of health care and understanding departmental responses to various changes are challenging. Simulation models, as suggested by system dynamics (SD) methodology, can help to increase this understanding by allowing scenarios to be tested in-silico before implementation. To this end, patient inflows to RT departments constitutes the first step of the RT process and must be thoroughly understood to create the initial parts of the model. Patients are treated with different intents for numerous diagnoses, typically leading to over 100 possible workflows for a large RT department. In this work, we investigate to what extent individual workflows can be reduced (aggregated) based on similarity in resource use to meet requirements of an SD-model where the aim is to keep the data input format small. Material and Methods We used real data for patients treated with curative and palliative intent at a seven-linac RT department in Sweden during 2015-2016. Workflow similarity was investigated by Mann-Whitney U tests and pair-wise correlations ( r ) between all possible combinations of workflows, with and without consideration of treatment intent. Similarity was quantified by averaged absolute pair-wise utility rate differences (%) and correlation analysis ( r ). Grouping of workflows was decided using two customized algorithms: 1. All elements correlate with one main element; 2. All elements correlate with each other. Both algorithms were applied to five correlation coefficient cutoffs ( r =0.75/0.80/0.85/0.90/0.95) to identify the smallest number of workflow groups. Results During the studied period, 128 workflows could be distinguished for 3209 patients (72 workflows for 2094 patients with curative intent/56 workflows for 1115 patients with palliative intent). Workflow dissimilarity was indicated for <1%; correlations were generally ≥0.87 (median). For grouping algorithm 1, median number of groups (maximum within-group differences) for workflows with curative intent only were in the range 10-33 (107- 279%), for palliative intent only 5-14 (37-63%), and without consideration of treatment intent 11-39 (30-123%). For grouping algorithm 2, corresponding numbers were 10-45 (41-279%), 5-39 (32-92%), and 11-87 (60-252%). Detailed results in Table 1.