Session Item

Adaptive radiotherapy requires rapid recontouring for replanning or assessment of need to replan. Studies have reported the geometric accuracy of propagated contours, but no studies have looked at the dosimetric implications of using these contours for replanning. In this work, propagated contours from 5 software applications were used to reoptimize head and neck (HN) radiotherapy plans on treatment CBCT images.

Radiotherapy treatment plans for ten randomly selected HN patients, using 65.1/54 Gy to PTV1/PTV2 in 30 fractions, were created in RayStation (RayStation Labs, Sweden). Planning contours for the spinal cord (SC), brainstem (BS), parotids and larynx contours were propagated to five CBCTs (equally spaced during treatment) which underwent a shading correction¹. Contours were propagated using 5 commercially available systems: Pinnacle (Philips, NL), Mirada (Mirada Medical Systems, UK), ProSoma (Medcom, FR), RayStation and ADMIRE (Elekta, UK). Two gold standard contours were created: manual and consensus from all propagated contours (STAPLE). PTVs were derived from rigidly propagated CTVs.

Plans based on each CBCT for all propagated contours (n=5 sets) were optimised on RayStation via the scripting interface. DVH parameters for each replan, using the gold standard contours, were extracted. Differences between the DVH parameters of the plans optimised using propagated contours and the plans optimised using the gold standard contours were assessed using Wilcoxon sign rank test. For reporting, parotids were split into spared and treated depending on whether the planned average dose was <26 Gy. To assess for plan complexity, the modulation complexity score (MCS)² and the plan MU for the replans were extracted.

A total of 350 replans were generated (10 patients, 7 contour sets, 5 CBCTs). Median DVH differences were maximum 0.4 Gy (table 1, fig 1), with maximum DVH difference outliers up to 7.3 Gy – possibly due to dose gradients. The median MU (fig 2a) for the original plans was 467.4 MU and 555.3 MU for the replans (p<0.01) whilst the MCS (fig 2b) was 0.127 and 0.119 (p=0.26) – indicating a slight increase in complexity.

Table 1. Median DVH parameter differences (Gy), D1cc for SC, BS and their PRVs, mean doses for parotid and larynx. * indicates statistically significant differences (p<0.05)

In our patient group, most propagated contours were able to produce a plan close to the gold standard. Contour correction would be recommended in high dose and high dose gradient organs. The replans calculated in this study have an increased complexity, so deliverability should be further assessed. Further work on patients with large changes is required to evaluate the generalisability of our conclusions.

¹Marchant, T.E. et al., Phys Med Biol, 2008, ²McGarry, C.K. et al, Br J Radiol, 2015.