Purpose: FDG PET is increasingly used in radiotherapy planning. Recently, we demonstrated substantial differences in target volumes when applying different methods of FDG-based contouring in primary lung tumours (Nestle et al., J Nucl Med 2005;46:1342-8). This paper focusses on FDG-positive mediastinal lymph nodes (LN(PET)).
Methods: In our institution, 51 NSCLC patients who were candidates for radiotherapy prospectively underwent staging FDG PET followed by a thoracic PET scan in the treatment position and a planning CT. Eleven of them had 32 distinguishable non-confluent mediastinal or hilar nodal FDG accumulations (LN(PET)). For these, sets of gross tumour volumes (GTVs) were generated at both acquisition times by four different PET-based contouring methods (visual: GTV(vis); 40% SUVmax: GTV40; SUV=2.5: GTV2.5; target/background (T/B) algorithm: GTV(bg)).
Results: All differences concerning GTV sizes were within the range of the resolution of the PET system. The detectability and technical delineability of the GTVs were significantly better in the late scans (e.g. p = 0.02 for diagnostic application of SUVmax = 2.5; p = 0.0001 for technical delineability by GTV2.5; p = 0.003 by GTV40), favouring the GTV(bg) method owing to satisfactory overall applicability and independence of GTVs from acquisition time. Compared with CT, the majority of PET-based GTVs were larger, probably owing to resolution effects, with a possible influence of lesion movements.
Conclusion: For nodal GTVs, different methods of contouring did not lead to clinically relevant differences in volumes. However, there were significant differences in technical delineability, especially after early acquisition. Overall, our data favour a late acquisition of FDG PET scans for radiotherapy planning, and the use of a T/B algorithm for GTV contouring.