18F-FDG-PET-based tumor delineation in cervical cancer: threshold contouring and lesion volumes.

OBJECTIVE To evaluate a semi-automated PET-image tumor segmentation algorithm for gross tumor volume (GTV) delineation in patients with locally advanced cervical cancer. MATERIAL AND METHODS Thirty-two patients with locally advanced cervical cancer were retrospectively evaluated. Semi-automated PET-image-based GTV delineation was applied using a previous established algorithm (GTV2SD) and 2 fixed threshold-based methods (GTV40% and GTV50%). GTV2SD was determined as the pixel with the mean value plus 2-standard deviation of the liver intensity, and GTV40% and GTV50% with 40% and 50% of the maximum tumor intensity (Tmax), respectively. The derived volumes were then compared with the GTVs generated manually using MR (GTVMR). RESULTS The mean value of GTV2SD, GTV40% and GTV50% was 85.3cc, 16.2cc and 24.1cc, respectively. Good agreement was noticed between GTV2SD and GTVMR (ρ=0.88). GTV40% and GTV50% showed weaker correlation with GTVMR (ρ=0.68 and ρ=0.71, respectively). CONCLUSIONS This study provides preliminary evidence that metabolic tumor volume delineation is feasible using computer-generated measurements in (18)F-FDG PET images. Generation of PET-based tumor volumes is affected by the choice of threshold level used. Metabolic tumor bulk calculated using the pixel with the mean value plus 2-standard deviations of the liver intensity (GTV2SD) correlates better with the MR-derived tumor volumes. The method is a simple and clinically applicable approach to generate PET-derived GTV for radiation therapy planning of cervical cancer.