ESTRO 2022 - Abstract Book

S1425

Abstract book

ESTRO 2022

Low-dose 4DCBCT has a minimal effect on automatic registration accuracy and tumour motion estimates at 30% and 10% of clinically used doses. Decreasing the dose to 30% did not have a considerable detrimental effect on image quality, since the impact of streak image artefacts exceed that of image noise at this dose level. This work highlights the feasibility of reduced-dose 4DCBCT, which could be relevant for patients for whom 4DCBCT is avoided due dose concerns. We show that using the same or lower dose than in clinical 3D modes is acceptable.

PO-1632 DL-based OAR delineation for Head and Neck Radiotherapy: accuracy versus computational resources

L. Cubero 1 , J. Serrano 2 , F.A. Calvo 2 , A. Simon 3 , J. Castelli 3 , R. De Crevoisier 3 , Ó. Acosta 3 , J. Pascau 1

1 Universidad Carlos III de Madrid, Bioengineering and Aeroespace Engineering - IGT, Madrid, Spain; 2 Clínica Universidad de Navarra, Departamento de Oncología Radioterápica, Madrid, Spain; 3 Université de Rennes I, CLCC Eugene Marquis, INSERM, LTSI-UMR 1099, Rennes, France Purpose or Objective Contouring the organs at risk (OAR) accurately in head and neck (HN) radiation therapy planning is crucial for reducing treatment-induced toxicity. These delineations are time and labor-consuming and often biased by inter and intra-observer variability. Automatic deep-learning (DL) based segmentation has proven to overcome the limitations of manual delineation, yielding more robust, patient-specific contours faster. Nonetheless, these algorithms have not been integrated into the radiotherapy workflow yet, mainly constrained by the need for extensive computational resources and technical experience. This study aims to compare two different DL algorithms and assess this technology's potential in HN radiotherapy. Materials and Methods 45 CT images from HN cancer patients with manually segmented OAR (brainstem, cord, eyes and parotids) were split into training (n = 35) and test (n = 10) sets. Two different fully convolutional neural networks were trained to segment the 4 OAR. On the one hand, the nnU-Net, a method that has shown great accuracy in anatomical delineation by automatic hyperparameter configuration and 5-fold cross-validation. This complex architecture leads, however, to long training and inference times. On the other hand, a single-class DenseVNet was trained for each OAR, using as input a bounding box built from a coarse mask obtained with a multiclass 3DUnet. This network presents certain architectural advantages that result in shorter training and inference. Both algorithms were evaluated in the test set by computing the Dice Score Coefficient (DSC) and Average Surface Distance (ASD). Results Figures 1 and 2 depict the evaluation of both DL algorithms on the test set. nnU-Net achieved slightly more accurate results for every OAR but required over 180 hours for training and 5 minutes for inference, both in an Nvidia RTX 8000 GPU. Instead, each instance of the DenseVNet was trained in approximately 1.5 hours on the same GPU, a total of 6 hours for all OAR, whereas the inference drops to around 70 seconds on CPU. Moreover, DenseVNet allows for retraining one class or introducing a new OAR independently, while nnU-Net must be retrained entirely.