ESTRO 2020 Abstract book

S205 ESTRO 2020

0.77±0.16(ABS19). When comparing ABS19 to ABS26 an increase of 10% was observed for the bladder contour (0.76±0.15 to 0.84±0.07). Despite this increase the contours using the MBS model were always better. Different results were obtained for the rectum and the femurs where increasing the number of patients in the model from ABS19 to ABS26 did not improved the DICE index: rectum (0.74±0.08 vs 0.73±0.08) and femurs (0.95±0.01) for both. On average, contours done using an ABS model take from 3-10 minutes to be created depending on the number of patients used to create the model compared to 1-2 minutes using MBS. Conclusion Despite the fact that our results show that MBS should be use for contouring the bladder and ABS for the rectum in our clinic we found that drawing these contours by hand gives better results in less time. For the femurs, since both models gave comparable results; we decided to use MBS based on the time it takes to create the contours compared to ABS. In the future we would like to do further comparisons of our findings vs. deep-learning in RS. OC-0354 Cautiously optimistic: A survey of radiation oncology professionals’ perceptions of automation V. Batumalai 1,2 , M.G. Jameson 1,2 , O. King 1 , R. Walker 1 , C. Slater 1 , K. Dundas 1,2 , G. Dinsdale 1 , A. Wallis 1 , C. Ochoa 1 , R. Gray 1 , P. Vial 1 , S.K. Vinod 1,2 1 South Western Sydney Local Health District, Department of Radiation Oncology & Ingham Institute, Sydney, Australia ; 2 University of New South Wales, South Western Sydney Clinical School, Sydney, Australia Purpose or Objective Given the increased demand for health services, automation processes and technological advances within the workforce are increasing. While there is evidence to show the positive effects of automation in improving overall radiotherapy department efficiency, there is no research to show how radiation oncology professionals perceive these changes. This study examined radiation oncology professionals’ perceptions of automation in An online survey link was sent to the chief radiation therapists of all Australian radiotherapy centres. It was requested that the survey be sent to all radiation therapists (RT), medical physicists (MP) and radiation oncologists (RO) within their institution. The survey was open from May – July 2019. Questions included the current and planned level of automation in departments and opinions on the effect of automation on specific tasks, roles and jobs. Results Participants were 204 RTs, 84 MPs and 37 ROs with estimated response rates of 10% of the overall Australian radiation oncology workforce. Respondents reported that most planning tasks are ‘somewhat automated’ or ‘automated with manual tuning’ (Figure 1). 69% of respondents felt very probably/probably empowered to drive decisions about implementing automated planning processes. 66% of respondents indicated they thought automation in planning was very important/important. Respondents felt automation resulted in improvement in work output and productivity (88%), quality of planning (57%), consistency in planning (90%) and staff focus on patient care (49%). When asked about perceived impact of automation, the following responses were recorded; will change the primary tasks of certain jobs (66%), will allow staff to do the remaining components of their job more radiotherapy planning. Material and Methods

Conclusion Using surface guided radiotherapy for ultra- hypofractionated prostate cancer patients reduced the patient setup time significantly with 42 seconds per treatment session. Furthermore, the initial setup accuracy was improved. OC-0353 Finding the contouring tool (manual vs automated segmentation) for prostate patients in RayStation F. Vallejo Castañeda 1 , M. Hinse 1 1 Centre intégré de cancérologie de Laval, Département de radio-oncologie, Laval, Canada Purpose or Objective To find the best contouring tool between manual-expert contours exported from Pinnacle vs automated contours done in RayStation (RS) using either an ATLAS base (ABS) or a model base (MBS) segmentation. Material and Methods A total number of 43 patients was used for the study. From them, 26 were used to construct 3 ABS models with a different specification and number of patients each (ABS15, ABS19 and ABS26). The remaining 17 patients were used for validation. Every validation patient was imported into RS and further sets of contours were created (ABS and MBS). Each ABS had expert contours of the bladder, rectum, and the femurs. ABS19 was the model applied to all validation patients. It was further used to compare it to ABS26 to test if increasing the number of patients to create a model would further improve the contours. ABS15 was the model for which a selection of patients with an specific bladder volume between 100-200cc was created. We wanted to test if the final contour of an organ could be improved by being in the same volume range as the model. From the validation patients; 7 were selected for which the volume of the bladder was in the range of ABS15. For them 2 further bladder contours were created using ABS19 and ABS15. Finally, the metrics used to compare our results were the mean DICE similarity coefficient between the expert vs. automated model and the time it took to run each model. Results For the bladder, MBS contours gave better results than ABS19 (0.91±0.06 vs 0.76±0.15). For the rectum the opposite was observed where ABS19 was better than MBS (0.72±0.08 vs 0.55±0.20). For the femurs comparable results were obtained with both models (0.95±0.01 (ABS19) vs 0.93±0.01 (MBS)). For the 7 patients selected to test ABS15, no significant increase in the mean DICE index for the bladder was obtained (0.75±0.11(ABS15) vs