ESTRO 2024 - Abstract Book

S4317

Physics - Intra-fraction motion management and real-time adaptive radiotherapy

ESTRO 2024

United Kingdom. 5 The Christie Hospitals NHS Foundation Trust, Medical Physics and Engineering, Manchester, United Kingdom

Purpose/Objective:

Respiratory motion is a challenge for delivering abdominal magnetic resonance (MR)- guided radiotherapy on an MR Linac, as it can impact image quality. Inaccuracies or variation in target and organs at risk (OAR) delineation due to image quality contribute significantly to radiotherapy workflow uncertainties. Abdominal compression (AC) and breath hold (BH) are two commonly used strategies to manage respiratory motion, however the impact of each strategy on interobserver variation in abdominal OAR contouring has not been investigated. This work evaluated the impact of these strategies compared to free-breathing (FB) on inter- and intra-observer variability of abdominal OAR delineation.

Material/Methods:

Participants recruited to an institutional imaging study (QUANTUM, NCT04748094 (1)) were included. At the imaging visit, 3D T2-weighted (W) turbo spin-echo (TSE) images were acquired in FB and with an AC belt. T1W 3D mDixon exhale images were acquired for BH due to shorter acquisition times, with the in-phase images selected for comparison.). Two of the included study participants had two imaging visits. All MRIs were anonymized, and the order randomized to minimize observer bias. Three OAR per scan were delineated in the clinical Monaco treatment planning system (v5.51.11) on MRIs by four expert observers (two radiation oncologists and two radiation therapists) over a period of approximately five months. Delineated OAR included stomach, liver, and duodenum. Observers were blinded to the motion management strategy used for each scan. A subset of scans, one scan per motion management strategy, were re-delineated by observers at a second time point (minimum of six weeks later) to calculate intra-observer variation. Observers were blinded to which scans were repeats. Simultaneous truth and performance level estimation (STAPLE) contours per OAR were generated using ADMIRE (Elekta AB, Stockholm, Sweden). These, along with observer contours, were imported to Raystation (v11B) for calculation of structure similarity metrics including Hausdorff distance (HD) and mean distance to agreement (mDTA). One participant had undergone a pancreaticoduodenectomy, resulting in omission of the duodenum contour on their cases, a mathematical weighting was applied for comparison of remaining cases with other organs. Statistical tests including one-way ANOVA were carried out to determine the statistical significance of difference in similarity metrics for contours.

Results:

Six participants were included in this study: one non-patient volunteer and five patients receiving abdominal radiotherapy. A total of 24 scans and 3 repeats over eight imaging visits were contoured by four observers. For contour similarity, stomach mDTA was 0.11 cm, 0.15 cm and 0.13 cm for AC, BH and FB respectively. Further results for individual organs are summarised in Table 1. No statistical significance was found for interobserver variation across motion management strategies using one-way ANOVA for duodenum, liver or stomach (p = 0.506, 0.971 and 0.577, respectively). Intra-observer variation was highest for all organs in FB (Fig. 1), most notably for the duodenum, with mDTA of 1.08 cm, 0.18 cm and 0.27 cm for duodenum, liver and stomach, respectively. Intra-observer mDTA for