External validation of deep learning-based contouring of head and neck organs at risk.

Ellen J L Brunenberg,René Monshouwer,Wouter Van Elmpt,Isabell K Steinseifer,Sven Van Den Bosch,Charlotte L Brouwer,Johannes H A M Kaanders,Mark J Gooding

doi:10.1016/j.phro.2020.06.006

Abstract

Background and purposeHead and neck (HN) radiotherapy can benefit from automatic delineation of tumor and surrounding organs because of the complex anatomy and the regular need for adaptation. The aim of this study was to assess the performance of a commercially available deep learning contouring (DLC) model on an external validation set.Materials and methodsThe CT-based DLC model, trained at the University Medical Center Groningen (UMCG), was applied to an independent set of 58 patients from the Radboud University Medical Center (RUMC). DLC results were compared to the RUMC manual reference using the Dice similarity coefficient (DSC) and 95th percentile of Hausdorff distance (HD95). Craniocaudal spatial information was added by calculating binned measures. In addition, a qualitative evaluation compared the acceptance of manual and DLC contours in both groups of observers.ResultsGood correspondence was shown for the mandible (DSC 0.90; HD95 3.6 mm). Performance was reasonable for the glandular OARs, brainstem and oral cavity (DSC 0.78–0.85, HD95 3.7–7.3 mm). The other aerodigestive tract OARs showed only moderate agreement (DSC 0.53–0.65, HD95 around 9 mm). The binned measures displayed the largest deviations caudally and/or cranially.ConclusionsThis study demonstrates that the DLC model can provide a reasonable starting point for delineation when applied to an independent patient cohort. The qualitative evaluation did not reveal large differences in the interpretation of contouring guidelines between RUMC and UMCG observers.

Highlights

In image-guided radiotherapy, the amount of data to be segmented is continuously expanding
An independent external validation of a deep learning contouring (DLC) model was performed on a set of 58 head and neck (HN) cancer patients
Reasonable and good correspondence was shown for the glandular OARs, mandible, brainstem and oral cavity, while the other aerodigestive tract OARs showed only moderate agreement

Summary

Introduction

In image-guided radiotherapy, the amount of data to be segmented is continuously expanding This is due to multi-modality imaging, adaptive radiotherapy, an increasing number of structures correlated with radiation-induced toxicity, and modern treatment modalities that improve organs-at-risk (OARs) sparing. Head and neck (HN) radiotherapy can benefit from automatic delineation of tumor and surrounding organs because of the complex anatomy and the regular need for adaptation. Materials and methods: The CT-based DLC model, trained at the University Medical Center Groningen (UMCG), was applied to an independent set of 58 patients from the Radboud University Medical Center (RUMC). DLC results were compared to the RUMC manual reference using the Dice similarity coefficient (DSC) and 95th percentile of Hausdorff distance (HD95). A qualitative evaluation compared the acceptance of manual and DLC contours in both groups of observers. The qualitative evaluation did not reveal large differences in the interpretation of contouring guidelines between RUMC and UMCG observers

Objectives

Methods

Results

Conclusion