Evaluating the impact of a rigid and a deformable registration method of pre-treatment images for hypoxia-based dose painting

•Rigid/deformable image registrations of FMISO-PET/CT to planning CT were compared.•The registration impacts on localisation and morphology of the hypoxic compartment.•The registration impacts the dose distribution of hypoxia-based dose painting.•Tumour control probability differences are negligible between registration methods. To assess the impact of rigid and deformable image registration methods (RIR, DIR) on the outcome of a hypoxia-based dose painting strategy. Thirty head and neck cancer patients were imaged with [18F]FMISO-PET/CT before radiotherapy. [18F]FMISO-PET/CT images were registered to the planning-CT by RIR or DIR. The [18F]FMISO uptake was converted into oxygen partial pressure (pO2) maps. Hypoxic Target Volumes were contoured on pO2 maps for the deformed (HTVdef) and non-deformed (HTV) cases. A dose escalation strategy by contours, aiming at 95 % tumour control probability (TCP), was applied. HTVs were characterised based on geometry-related metrics, the underlying pO2 distribution, and the dose boost level. A dosimetric and radiobiological evaluation of selected treatment plans made considering RIR and DIR was performed. Moreover, the TCP of the RIR dose distribution was evaluated when considering the deformed [18F]FMISO-PET image as an indicator of the actual target radiosensitivity to determine the potential impact of an unalignment. Statistically significant differences were found between HTV and HTVdef for volume-based metrics and underlying pO2 distribution. Eight out of nine treatment plans for HTV and HTVdef showed differences on the level 10 %/3 mm on a gamma analysis. The TCP difference, however, between RIR and the case when the RIR dose distribution was used with the deformed radiosensitivity map was below 2 pp. Although the choice of the CTplan-to-PET registration method pre-treatment impacts the HTV localisation and morphology and the corresponding dose distribution, it negligibly affects the TCP in the proposed dose escalation strategy by contours.