ESTRO 2020 Abstract Book

S344 ESTRO 2020

Mean perturbation, 1mm threshold percentage and mean percentage volume change were calculated per structure. Pointwise mean and SD of distance from raw contours were calculated and t-tests were carried out with a null hypothesis of 0mm edits. Qualitative visual analysis of rendered structures, displaying statistics on a reference surface highlighted systematic edits, occupying areas with large distance magnitude and low SD, and clusters of p -values < 0.01, suggesting that most clinicians made similar edits. Areas with high SD on rendered images may indicate interobserver variability. Results The spinal cord received the most editing (12.55 ± 6.04 mm) due to large superior-inferior displacements at the extremes; the cerebellum received the least (0.64 ± 0.44 mm) – measured by mean perturbation (Table 1). Excluding tubular structures, all structures received a mean perturbation of < 2.8 mm editing. However, summary statistics provide no information on the location of edits needed to more precisely identify systematic changes. Figure 1 shows localised information on renderings of the oral cavity and left parotid. Areas of systematic edits can be observed on the anterior-superior surface of the oral cavity (Figures 1a, 1b & 1c) and the extension of the parotid lobe (Figures 1d, 1e & 1f). Other areas of systematic edits were observed but are not displayed for reasons of space.

Conclusion Automated DL-based contour QA is feasible but visual inspection remains essential. We had a substantial number of false positive flags, due to sub-optimal performance of the DL model, especially for unusual anatomical deviations and deviating CT slice thickness. DL model improvement to better handle “outlier” cases will facilitate the adoption of DL-based contour QA into clinical trial QA and routine clinical practice. PH-0608 Identifying systematic edits in the clinical use of Deep Learning Contours J. Mateo 1 , P. Aljabar 1 , C. Brouwer 2 , S. Both 2 , M. Gooding 1 , H. Langendijk 2 1 Mirada Medical, Science, Oxford, United Kingdom ; 2 University Medical Centre Groningen, Radiotherapy, Groningen, The Netherlands Purpose or Objective To identify systematic edits of Deep-Learning Contouring (DLC) auto-contours of organs at risk (OAR) structures in a CT images were acquired for 77 head and neck cancer patients undergoing radiation therapy (RT) at the University Medical Centre Groningen. Contours for 21 OARs were obtained using an automated deep learning model (DLC Expert TM , Mirada Medical Ltd). Auto-contours were edited by clinical experts to ensure clinical validity for treatment planning, edited and raw (unedited) structures were then analysed. Not all OAR were present in all cases, resulting in variable counts for each structure. Edited structures were converted into meshes with scalar data representing the surface displacement to the corresponding raw contours. After rigid registration, and scaling to reference structures using an Iterative Closest Point algorithm, structures were re-meshed to a common topology with point correspondence. head and neck model. Material and Methods