Geometric and dosimetric evaluations of atlas-based segmentation methods of MR images in the head and neck region

J P Kieselmann,U Oelfke,C D Fuller,M J Menten,M J Cardoso,N Burgos,C P Kamerling,S Nill

doi:10.1088/1361-6560/aacb65

Abstract

Owing to its excellent soft-tissue contrast, magnetic resonance (MR) imaging has found an increased application in radiation therapy (RT). By harnessing these properties for treatment planning, automated segmentation methods can alleviate the manual workload burden to the clinical workflow.We investigated atlas-based segmentation methods of organs at risk (OARs) in the head and neck (H&N) region using one approach that selected the most similar atlas from a library of segmented images and two multi-atlas approaches. The latter were based on weighted majority voting and an iterative atlas-fusion approach called STEPS. We built the atlas library from pre-treatment T1-weighted MR images of 12 patients with manual contours of the parotids, spinal cord and mandible, delineated by a clinician. Following a leave-one-out cross-validation strategy, we measured the geometric accuracy by calculating Dice similarity coefficients (DSC), standard and 95% Hausdorff distances (HD and HD95), and the mean surface distance (MSD), whereby the manual contours served as the gold standard. To benchmark the algorithm, we determined the inter-observer variability (IOV) between three observers.To investigate the dosimetric effect of segmentation inaccuracies, we implemented an auto-planning strategy within the treatment planning system Monaco (Elekta AB, Stockholm, Sweden). For each set of auto-segmented OARs, we generated a plan for a 9-beam step and shoot intensity modulated RT treatment, designed according to our institution’s clinical H&N protocol. Superimposing the dose distributions on the gold standard OARs, we calculated dose differences to OARs caused by delineation differences between auto-segmented and gold standard OARs. We investigated the correlations between geometric and dosimetric differences.The mean DSC was larger than 0.8 and the mean MSD smaller than 2 mm for the multi-atlas approaches, resulting in a geometric accuracy comparable to previously published results and within the range of the IOV. While dosimetric differences could be as large as 23% of the clinical goal, treatment plans fulfilled all imposed clinical goals for the gold standard OARs. Correlations between geometric and dosimetric measures were low with R2 < 0.5.The geometric accuracy and the ability to achieve clinically acceptable treatment plans indicate the suitability of using atlas-based contours for RT treatment planning purposes. The low correlations between geometric and dosimetric measures suggest that geometric measures alone are not sufficient to predict the dosimetric impact of segmentation inaccuracies on treatment planning for the data utilised in this study.

Highlights

Magnetic resonance (MR) imaging has found an increased application in image guidance for radiation therapy (RT) owing to its superior soft-tissue contrast and lack of ionising radiation compared to the conventionally used x-ray computed tomography (CT) (Metcalfe et al 2013, Dirix et al 2014, Lagendijk et al 2014)
We first performed three different atlas-based segmentation methods using a library of manually segmented MR images, which is illustrated in the top part of figure 1
The only difference between approach A and the approaches B and C in terms of the computation time was attributed to the atlas selection and fusion method

Summary

Introduction

Magnetic resonance (MR) imaging has found an increased application in image guidance for radiation therapy (RT) owing to its superior soft-tissue contrast and lack of ionising radiation compared to the conventionally used x-ray computed tomography (CT) (Metcalfe et al 2013, Dirix et al 2014, Lagendijk et al 2014). Treatment planning and dose calculation are solely based on the MR image but are challenging as the required electron density information cannot be derived directly from image intensities Methods such as creating synthetic CTs are necessary to provide surrogates for electron densities (Edmund and Nyholm 2017). Clinicians conventionally delineate all VOIs prior to treatment This is especially tedious for the treatment of head and neck (H&N) cancer patients due to the complex anatomy including many OARs and target volumes. Many of these VOIs are difficult to delineate on a CT and would benefit from MR imaging (Schmidt and Payne 2015). The most commonly used automated delineation methods are atlas-based (Fritscher et al (2014) and references therein)

Methods

Results

Discussion

Conclusion