Purpose: Delineating tumor motion by four-dimensional positron emission tomography/computed tomography (4D-PET/CT) is a crucial step for gated radiotherapy (RT). This article quantitatively evaluates semiautomatic algorithms for tumor shift estimation in the lung region due to patient respiration by 4D-PET/CT, in order to support the selection of the best phases for gated RT, by considering the most stable phases of the breathing cycle.
Methods: Three mobile spheres and ten selected lesions were included in this study. 4D-PET/CT data were reconstructed and classified into six/ten phases. The semiautomatic algorithms required the generation of single sets of images representative of the full target motion, used as masks for segmenting the phases. For 4D-CT, a pre-established HU range was used, whereas three thresholds (100%, 80%, and 40%) were evaluated for 4D-PET. By using these segmentations, the authors estimated the lesion motion from the shifting centroids, and the phases with the least motion were also deduced including the phases with a curve slope less than 2 mm/ delta phase. The proposed algorithms were validated by comparing the results to those generated entirely by manual contouring.
Results: In the phantom study, the mean difference between the manual contour and the semiautomatic technique was 0.1 +/- 0.1 mm for 4D-CT and 0.2 +/- 0.1 mm for the 4D-PET based on 40% threshold. In the patients' series, the mean difference was 0.9 +/- 0.6 mm for 4D-CT and 0.8 +/- 0.2 mm for the 4D-PET based on 40% threshold.
Conclusions: Estimation of lesion motion by the proposed semiautomatic algorithm can be used to evaluate tumor motion due to breathing.