Deformation of planned dose distribution to track delivered dose to Non Small Cell Lung Cancer (NSCLC) patients

Orban de Xivry, Jonathan [UCL] Janssens, Guillaume [UCL] Bosmans, Geert [] Buijsen, Jeroen [] De Craene, Mathieu [UCL] Dekker, A [] Van Baardwijk, Angela [] Lambin, Philippe [UCL] Macq, Benoît [UCL] et al.[show all]

Purpose/Objectif: Classic dose planning is performed on a single phase CT scan for lung cancer patients and does not take into account all respiratory deformations. The computation of accumulated delivered dose incorporating these deformations is therefore a challenging issue to improve the planning accuracy. Respiration-Correlated-CT (RC-CT) is a promising modality for quantifying the movement and deformation of the tumour and incorporating this information in treatment planning. We propose a method to deform dose distribution in order to track the dose delivered to each voxel through all respiratory phases. Materials/Methods: Dense displacement fields were computed using a combination of rigid and non-rigid registrations between RC-CT phases. The non-rigid registration uses a b-splines deformation model (on a regular grid with 15 mm spacing). The cross-correlation metric has been used as similarity measure both for rigid and non-rigid registration. These registrations have been performed independently between each phase and the reference phase. Dose distributions at every phase were deformed according to the displacement fields. These distributions were then added on the reference phase to see the total dose received by each voxel through the whole respiratory cycle. Results:The registration process was applied tO 11 different patients and was validated at each phase.This was carried out by comparing the overlap between a manual segmentation and an automatic segmentation resulting from the deformation of the delineation on the reference phase, according to the computed deformation field. The average concordance index for all patients reached 0.79, which is not significantly different from the inter-observer variability, showing that the registration is accurate. The dose deformation process was then validated, currently on 6 of the 11 patients : a histogram of the deformed dose distribution inside the tumour on the reference phase was computed and compared to the histogram of the dose distribution inside the tumour for each phase. In the case of large movements, the correlation between the two histograms was significantly better after deformation (p-value < .05). Conclusions: We developed a method to use RC-CT information into the radiation treatment planning. This method can also be applied in the context of dose-guided radiotherapy, by adjusting the dose planning according to the deformation between a pre-treatment CT-scan and a CT-scan during treatment.