Evaluation of different methods of 18F-FDG-PET target volume delineation in the radiotherapy of head and neck cancer.

OBJECTIVE: To quantify differences between the alternative methods of F-fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET)-based delineation of the gross tumor volume in patients with head and neck cancer.

MATERIALS AND METHODS: Twelve patients with locally-advanced head and neck carcinomas were studied. The reference gross tumor volume (GTVref) was established by a radiation oncologist, along with a neuroradiologist, using the computed tomography-simulation and diagnostic magnetic resonance imaging data. With the GTVref obscured, a second radiation oncologist and a nuclear medicine physician determined the following contours: (1) high FDG uptake based on visual inspection (GTVvis), (2) the contour derived from the 50% maximum standardized uptake value (SUV) threshold (GTV50), (3) the contour derived from a 2.5 SUV absolute threshold (GTV2.5), and (4) the contours derived from an iterative segmentation algorithm (GTViter). These volumes were compared with the GTVref using a signed-ranks test with the exact reference distribution.

RESULTS: The average GTVref was 75.5 mL (median 72.8 mL, range 22.2-138.4 mL). The average GTVvis was 57.6 (median 55.4 mL, range 12-115.8 mL). Overall, a 21% reduction in volume size was observed with GTVvis versus GTVref. When the signed-ranks test with the exact reference distribution was applied, the difference was not statistically significant (P = 0.32). The average GTV2.5 was 60 mL (median 64.5, range 8.8-90.3 mL). The differences between GTV2.5 and GTVref were not statistically significant (P = 0.35). The use of GTV50 and GTViter produced significantly smaller volumes with respect to GTVref (P < 0.005).

CONCLUSIONS: PET-based tumor volumes are strongly affected by the choice of threshold level. Quantitatively, GTVs derived from visual inspection of the region of high FDG uptake do not significantly differ from GTVref in this cohort of patients. The inclusion of alternative FDG-PET segmentation data, other than visual inspection, may reduce target volumes significantly.