ESTRO 2020 Abstract book

S372 ESTRO 2020

HKNPCSG, NCIC CTG, NCRI, NRG Oncology and TROG consensus guidelines. Radiotherapy & Oncology 2015.

PH-0606 AutoConfidence: Per-patient validation for clinical confidence in deep learning for radiotherapy M. Nix 1 , D. Bird 1 , M. Tyyger 1 , A. Appelt 1 , L. Murray 1 , H. McCallum 2 , B. Al-Qaisieh 1 , A. Gooya 3 1 St James Institute of Oncology, Radiotherapy Physics, Leeds, United Kingdom ; 2 Newcastle University Teaching Hospitals, Medical Physics, Newcastle, United Kingdom ; 3 University of Leeds, Computer Science, Leeds, United Kingdom Purpose or Objective Deep-learning (DL) has proven potentially powerful for auto-contouring (AC), synthetic CT (sCT) generation from MRI. Commissioning and validation for DL methods are extremely challenging due to their ‘black-box’ nature and training-dependent robustness to variations in input data. Errors and uncertainties are difficult to quantify and vary inter- and intra-case. We demonstrate a conditional generative-adversarial-network (cGAN), as the core of a robust AI strategy, dubbed AutoConfidence, which is capable of identifying outliers in unseen input data, and locally assessing the uncertainties and errors in DL predictions (contours or sCT). Crucially, error and uncertainty analysis can be decoupled from the generation network (fig. 1), allowing independent per-patient validation of DL-based sCT or auto-contours from any source, including CE/FDA approved systems. Material and Methods 32 t2w-SPACE RT position pelvic MRs (2240 slices), from anorectal cancer patients, were deformably registered to planning CT and used to adversarially train a cGAN (fig. 1), consisting of a generative U-net and separate discriminative shallow U-net in an extension of the popular style-transfer network pix2pix . The discriminative network was trained to predict Hounsfield unit error in the generator output relative to reference CT. Auto- contouring: Planning CT from 16 prostate RT patients (1520 slices), with expert clinical contours, were used to train the cGAN to generate and quality assess contours for 7 OARs. The discriminative network was trained to predict local misclassification relative to reference contours. In both cases single-class support vector machine (1-SVM) and auto-encoder outlier-detection scoring allowed detection of unseen input images lying beyond the confidence limits of the trained networks. Per-slice confidence scores were produced alongside local confidence maps, providing metrics for acceptance/rejection and confidence maps for human intervention/editing.

Conclusion This external validation study proves that the used DLC model can be applied to a new patient cohort, provided that delineation standards are sufficiently similar. The binned DSC gives local information that is very useful to identify deviations, either in delineation, in patient anatomy, or due to slice thickness/partial volume issues. To further evaluate these deviations and the clinical acceptance of the results, follow-up research will focus on clinical evaluation of the deep learning contouring. References 1. Van Dijk L, et al. Improving automatic delineation for head and neck organs at risk by Deep Learning Contouring. Radiotherapy & Oncology 2019. 2. Brouwer C, et al. CT-based delineation of organs at risk in the head and neck region: DAHANCA, EORTC, GORTEC,