# Standardising image registration and dose mapping for thoracic reirradiation: A national multi-centre benchmarking study

## TL;DR

This study compares rigid and deformable image registration methods for lung cancer reirradiation, finding that deformable methods improve consistency and reduce dose variability.

## Contribution

Demonstrates that deformable image registration (DIR) reduces dose variability compared to rigid methods in multi-centre lung reirradiation settings.

## Key findings

- DIR significantly reduced geometrical variations for organs-at-risk with mean surface distances under 4.4 mm.
- DIR showed less variability in near-maximum EQD2-doses compared to rigid registration.
- No significant differences were found in full DVH bands between registration methods.

## Abstract

•Lung reirradiation dose accumulation across six centres.•Image registration performed using three software solutions.•DIR (deformable) improved organ-at-risk alignment compared to RIR (rigid) methods.•Applying CURE Lung constraints DIR gave statistically less dose variation than RIR.•Findings support standardised use of DIR for safe lung cancer reirradiation.

Lung reirradiation dose accumulation across six centres.

Image registration performed using three software solutions.

DIR (deformable) improved organ-at-risk alignment compared to RIR (rigid) methods.

Applying CURE Lung constraints DIR gave statistically less dose variation than RIR.

Findings support standardised use of DIR for safe lung cancer reirradiation.

Safe delivery of high-dose reirradiation depends on robust evaluation of previously delivered dose on current patient anatomy. This study aims to assess the variability of 1) rigid (RIR) and deformable (DIR) image registration, and 2) dose-mapping based on RIR or DIR, as part of the pre-trial quality assurance for CURE Lung, a Scandinavian thoracic reirradiation trial.

Seven lung cancer cases, treated with curatively intended reirradiation near a previously irradiated region, were distributed to six Danish centres. Organs-at-risk (OARs) were delineated on CT images of previous (CTprev) and current treatment (CTcurrent). RIR and DIR from CTprev to CTcurrent were performed for all cases at each centre, and doses were mapped from CTprev to CTcurrent using both methods. Mapped and current physical doses were converted to equieffective doses (EQD2), and cumulative EQD2-doses were generated on CTcurrent. Inter-centre variation was assessed using volume coefficients and distance measurements. A linear mixed-effects model tested the effect of registration method on OAR alignment, near-maximum and volumetric EQD2-dose constraints from CURE Lung, and variability in DVH bands.

DIR significantly reduced the geometrical variations for all OARs, with mean surface distances <4.4 mm. Near-maximum EQD2-doses overall had less variability with DIR compared to RIR. No significant differences were found for the full DVH bands.

DIR provides significantly better consistency in transfer of OAR from CTprev to CTcurrent and reduces variability in EQD2-dose evaluations compared to RIR. This supports preferential use of DIR for reirradiation planning in lung cancer.