Incorporation of motion information for tumour volume definition in PET

The interest of PET complementary information for the delineation of the target volume in radiotherapy of lung cancer is increasing. However, respiratory motion requires the determination of a functional internal target volume (ITV) on PET images for which several strategies have been proposed. The purpose of this study was the comparison of these strategies for taking into account respiratory motion and deriving the ITV: 1) adding fixed margins to the volume defined on a single binned image, 2) segmenting a motion averaged image, and 3) considering the union of volumes delineated on binned frames. For this third strategy, binned frames were either non-corrected for motion, or corrected using two different methods: elastic registration or super resolution. The strategies' performances were assessed on realistic simulated datasets combining the NCAT phantom with a PET Philips GEMINI scanner model in GATE, and containing various configurations of tumor to background contrast, with both regular and irregular respiratory motion (with a range of motion amplitudes). The obtained ITVs' sensitivity (SE) and positive predictive value (PVE) with respect to the known true ITV were significantly higher (from 0.8 to 0.95) than all other techniques when using binned frames corrected for motion, independently of motion regularity, amplitude, or tumor to background contrast. Although the absolute difference was small and not always significant, images corrected using super resolution led to systematically better results than using elastic registration. The worst results were obtained when using the motion averaged image for sensitivity (around 0.5-0.6) and using the margins added to a single frame for PPV (0.6-0.7) respectively. The best strategy to account for breathing motion for tumor ITV delineation in radiotherapy planning is to rely on the use of the union of volumes delineated on super resolution-corrected binned images. 1. Introduction Computed tomography (CT) is routinely used for delineating gross target volumes (GTV) and organs at risk (OAR) in radiotherapy treatment planning (RTP) of lung cancer. However, the high sensitivity and specificity of 18 F-FDG Positron Emission Tomography (PET) for diagnosis and staging, as well as the widespread availability of combined PET/CT scanners, led to numerous studies investigating the potential of PET/CT for target volume delineation. Use of PET/CT images has been shown to reduce intra and inter observer variability in target volume delineation [Ashamalla 2005, Steenbakkers 2006, van Baardwijk 2007], as well as significantly modifying GTVs [Walter 1999, Bradley 2004]. Moreover, PET imaging allows characterizing tracer uptake heterogeneities than …