ESTRO 2020

Session Item

November 28
10:30 - 11:30
Physics Stream 1
Proffered papers 5: Analysis for toxicity and outcome
Proffered Papers
10:50 - 11:00
Voxelwise analysis of rectal toxicity associations using accumulated delivered dose
Leila Shelley, United Kingdom


Voxelwise analysis of rectal toxicity associations using accumulated delivered dose
Authors: Amy Bates.(Addenbrooke's Hospital, Cambridge Clinical Trials Unit, Cambridge, United Kingdom), Neil Burnet.(University of Manchester, Manchester Academic Health Science Centre, Manchester, United Kingdom), Karl Harrison.(University of Cambridge, Cavendish Laboratory, Cambridge, United Kingdom), Raj Jena.(Addenbrooke's Hospital, Oncology Centre, Cambridge, United Kingdom), David Noble.(Addenbrooke's Hospital, Oncology Centre, Cambridge, United Kingdom), Marina Romanchikova.(National Physical Laboratory, Radiotherapy, Teddington, United Kingdom), Leila Shelley.(Edinburgh Cancer Centre, Oncology Physics, Edinburgh, United Kingdom), Michael Sutcliffe.(University of Cambridge, Department of Engineering, Cambridge, United Kingdom), Simon Thomas.(Addenbrooke's Hospital, Medical Physics & Clinical Engineering, Cambridge, United Kingdom)
Show Affiliations
Purpose or Objective

Rectal toxicity remains a clinical issue in prostate radiotherapy. Associations between dose and toxicity are generally poorly understood. DVHs are commonly used to quantify dose to the rectum but have been criticised for lacking in geometric information. By evaluating spatial dose distribution to the rectal wall, stronger dose-toxicity associations may be revealed. Voxel-level analysis may indicate regions of intraorgan radiosensitivity.  Furthermore, calculating motion-inclusive delivered dose to the rectal wall may improve the power of predictive models.

Here we compare 3D dose distributions for patients with and without toxicity to identify rectal subregions at risk (SRR) for 12 endpoints. The discriminatory ability of planned and delivered dose is compared.

Material and Methods

Dose to the rectal wall was calculated using biomechanical finite element modelling (Abaqus, Dassault Systèmes) for 139 prostate cancer patients.  The 3D anatomical model was grown and deformed based on rectal contours either manually defined on the planning scan, or autosegmented on daily MVCT IGRT scans. Doses were calculated using CheckTomo.  Dose-histories were tracked per finite element, or voxel. Patients were treated with helical IGRT and prescribed 74 Gy/37# or 60 Gy/20# (combined by converting to EQD in 37#). Total delivered dose incorporates interfraction motion and was determined by accumulating daily dose.

Toxicity data were prospectively collected and cumulative incidence at 2 years post-treatment were investigated. Endpoints were: diarrhoea (26%), faecal incontinence (17 %), proctitis ≥G1 (19%) ≥G2 (14%), rectal bleeding ≥G1 (33%) ≥G2 (12%), GI toxicity ≥G1 (53%) ≥G2 (15%),  stool frequency (25%), bowel bother ≥G1 (40%) ≥G2 (23%), ≥G3 (9%). Patients were split by those with and without toxicity and a voxel-wise student’s t-test was performed to identify SRRs with p<0.05 (Fig 1). Post-processing involved removing small clusters, applying a smoothing filter and filling in holes. Equivalent uniform dose (EUD) was calculated from voxels within the resulting SRR, for both planned and delivered dose. The area under the receiver operator characteristic curve (AUC) was used to indicate discriminative ability.


SRRs differed between endpoints but were all located away from the high dose region adjacent to the prostate. Accumulated dose produced stronger associations than planned dose for 9/12 endpoints (Fig 2). However, differences could not be considered significant.


Voxel-level analysis of dose to the rectal wall revealed SRRs associated with rectal toxicity. Differences between planned and delivered dose associations were small. Results may improve understanding of the pathophysiology and radiosensitivity behind radiation-induced side-effects.