An Evaluation of the Reproducibility of Radiotherapy Contouring Utilizing Multiple Institutions and Treatment Planning Systems
E Pogson1,3*, J McNamara2, M Jameson1,3, R McDowall4, A Lim5, C Dempsey6, P Metcalfe1,3, L Holloway1,3,7,8, (1) University of Wollongong, Wollongong, NSW, (2) Illawarra Cancer Care Centre, Wollongong, NSW, (3) Liverpool and Macarthur Cancer Therapy Centres and the Ingham Institute, Liverpool, NSW, (4) Royal Brisbane and Womens Hospital, Brisbane, Queensland, (5) Peter MacCallum Cancer Centre, Melbourne, Victoria, (6) Calvary Mater Newcastle Hospital, Newcastle, NSW, (7) School of Physics, University of Sydney, NSW, (8) South Western Sydney Clinical School, University of New South Wales, NSW, AustraliaSU-E-J-213 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: Consistency of radiotherapy contours is required to ensure consistency of treatment and for this reason many studies have been undertaken and are expected in the future comparing contours of multiple observers or systems. This study was undertaken to determine the minimum uncertainty achievable when undertaking this type of investigation including multiple centres and treatment planning systems.
Methods: A Computed Tomography (CT) scan was taken of a commercially available uniformity Phantom. This dataset was then imported into various contouring software programs including Pinnacle, Xio and Focal at the same institution and variations at different institutions. Contours of the perimeter of the phantom and a detailed cylinder inside the phantom were contoured using the same observer at provided window levels. The perimeter of the phantom was auto-contoured using auto-threshold. The inside circle was contoured manually. Contours were then exported from the treatment planning systems and into CERR for analysis.
Results: A comparison of the phantom perimeter from Focal and Pinnacle at a single institution demonstrated a Concordance Index (CI) of 0.98, while the manually contoured cylinder has a CI of 0.77. When comparing between institutions the CI ranged from 0.75-0.85 for the cylinder. Variation in the phantom perimeter contours was mainly in the Z direction with 2 slices (0.4cm) not being contoured in Focal compared to Pinnacle. Maximum variation in the X and Y direction for the phantom perimeter was 0.098cm. The centre of mass of all phantom perimeter contours were within 0.10cm, with the largest variance between institutions occurring in the anterior-posterior direction.
Conclusion: The variation between auto-contouring and manually contouring a high contrast object for different treatment planning systems has been established. As expected manually contouring produces greater variation than auto-threshold contours between different treatment planning system.
Funding Support, Disclosures, and Conflict of Interest: Funding from Cancer Australia and The National Breast Cancer Foundation, Project grant 1033237