Interobserver variability in delineation of target volumes in head and neck cancer

Abstract

Background and purposeIn the last decade precision of radiotherapy treatment execution increased, demanding more accurate delineations to fully exploit these developments. The aim of this study was to identify the extent of interobserver variability in delineation of head and neck cancer (HNC). Materials and methodsIn February 2017 all Belgian radiotherapy departments were invited to complete an online survey and submit clinical target volumes for five HNC reference cases. Clinical target volume of the primary tumour (CTVp) and elective nodal neck (CTVe) were submitted and compared between centres for CTVp and to the CTVe ‘gold standard’ (CTVeGS). Volume, DICE similarity coefficient (DSC) and median Hausdorff Distance (HD) were measured and calculated. ResultsFourteen of 22 centres (64%) completed both survey and delineations. They all used delineation guidelines for CTVe and twelve confirmed the use of guidelines of Grégoire et al. Nine centres used CTVp guidelines, although none used the same ones. Median DSC for CTVe comparing centres with CTVeGS ranged between 0.67 and 0.82 and HD50 between 1.7 mm and 2.8 mm. Good agreement was shown for neck level II, III and IV, whilst worst consensus was observed for level Ib, V, VI, VIIa and VIIb. Improvement of DSC and HD50 was observed when the same levels as CTVeGS were selected. Median DSC and HD50 for CTVp ranged between 0.51 and 0.79 and 2.8 mm and 4.1 mm respectively, which both slightly improved when calculating it for only the centres using a 10 mm margin. ConclusionAlthough nearly all participants used identical guidelines for CTVe there were large discrepancies in neck levels selected and volumes delineated. CTVp delineations were also heterogeneous although we expect improvement with implementation of recently published guidelines. Additional teaching in target volume delineation is necessary as this paper demonstrates that availability and implementation of guidelines alone is not enough to guarantee uniform delineation.