| dc.description.abstract | Purpose: Radiotherapy (RT) of rectal cancer inevitably involves irradiation of the healthy bowel due to its close proximity. Reduction of excessive dose to the bowelis part of the radiation treatment plan optimisation. However, bowel position can vary during treatment due to bowel motion and the planned dose may not be representative for the dose delivered. Inaccurate delineation during planning increases uncertainties during RT delivery. This project aimed to test an algorithm for automatic bowel segmentation (auto-segmentation), and to study bowel motion during (intrafractional) and in between (interfractional) fractions. We also examined the impact of intrafractional bowel motion and variations in bladder filling on the dose distribution in the bowel during online adaptive radiotherapy (ART) of rectal cancer. Further, we investigated the impact of intrafractional bowel motion on risk estimates for grade ≥ 3 acute bowel toxicity.
Methods: In 15 rectal cancer patients, Cone-Beam Computed Tomography (CBCT) scans were acquired before (pre) and after (post) treatment for each of the first five fractions. The bladder, the individual small bowel and colon loops and the bowel (small bowel and colon combined) were segmented on each pre- and post-CBCT for every first (fx1) and last (fx5) fraction using a software for automatic segmentation, TotalSegmentator (TS), and by two manual delineators. Segmentation performances, intra- and interfractional bowel motion were quantified using the Dice Similarity Coefficient (DSC) and 95th Hausdorff Distance (HD95). The impact of inaccurate auto-segmentations and intrafractional bowel motion on the dose to the bowel were qualitatively and quantitatively investigated using dose-volume histogram (DVH) analyses for adaptive treatments. The relation between bladder volume and bowel exposure was investigated for low and high dose-volumes using statistical analyses. Dose-volume data considering intrafractional bowel motion were used in toxicity risk estimation for grade ≥ 3 acute bowel toxicity using two dose-response models.
Results: Statistically significantly poorer agreements were found between the bowel segmentations by TS and manual delineations (median DSC: 0.83, median HD95: 12.50 mm) than between the manually segmentations (median DSC: 0.90, median HD95: 7.9 mm) in post-CBCTs of fx1. The low dose-volumes for manual bowel segmentations were statistically significantly greater than the low dose-volumes for auto-segmented bowels. The intrafractional bowel motion in fx1 and fx5 (median DSC:0.83/0.84, median HD95: 12.59 mm/13.00 mm for fx1/fx5) was statistically significantly less than the interfractional bowel motion (median DSC: 0.72, median HD95: 20.97 mm). Intrafractional bowel motion had greater impact on low dose volumes than high dose volumes. Intrafractional increase in bladder volume was statistically significantly correlated with decrease in bowel exposure for volumes exposed to at least 20.0 Gy. The number of patients with risk estimates ≥ 5% for grade ≥ 3 acute bowel toxicity varied depending on the dose-response model.
Conclusions: TS did not segment the bowel in CBCTs with adequate performance, even when post-processed. Manual segmentations were able to measure intra- and interfractional bowel motion in CBCTs. Intrafractional bowel motion and increased bladder volume reduced bowel exposure. The risk estimates for grade ≥ 3 acute bowel toxicity including intrafractional motion showed substantial variations and require further investigation with treatment specific dose-response models. | |
