Delineation of small mobile tumours with FDG-PET/CT in comparison to pathology in breast cancer patients.

Abstract

Various segmentation methods for 18F-fluoro-2-deoxy-d-glucose (FDG) positron emission tomography/computed tomography (PET/CT) images were correlated with pathological volume in breast cancer patients as a model of small mobile tumours. Thirty women with T2-T3/M0 breast invasive ductal carcinoma (IDC) were included prospectively. A FDG-PET/CT was acquired 4 ± 3d before surgery in prone and supine positions, with/without respiratory gating. The segmentation methods were as follows: manual (Vm), relative (Vt%) and adaptive (Va) standard uptake value (SUV) threshold and semi-automatic on CT (Vct). Pathological volumes (Vpath) were measured for 26 lesions. The mean (±SD) Vpath was 4.1 ± 2.9 mL, and the lesion displacements were 3.9 ± 2.8 mm (median value: 3 mm). The delineated VOIs did not vary with the acquisition position nor with respiration, regardless of the segmentation method. The Vm, Va, Vct and Vt% methods, except Vt30%, were correlated with Vpath (0.5<r<0.8). The Vt50% and Vm were the most accurate methods (mean±SD: 0.0 ± 1.6 mL and +0.6 ± 1.8 mL, respectively), followed by the Vct method. When compared with pathology, small lesions (diameter <50mm) with limited respiratory displacement (i.e., breast or apical lung lesions) are best delineated on FDG-PET/CT using a 50% SUVmax threshold. The acquisition position and respiratory gating did not modify the delineated volumes.