Tumor volume delineation in head and neck cancer with 18-fluor-fluorodeoxiglucose positron emission tomography: adaptive thresholding method applied to primary tumors and metastatic lymph nodes

Abstract

There are several potential advantages of using 18-fluor-fluorodeoxiglucose (18F-FDG) PET for target volume contouring, but before PET-based gross tumor volumes (GTVs) can reliably and reproducibly be incorporated into high-precision radiotherapy planning, operator-independent segmentation tools have to be developed and validated. The purpose of the present work was to apply the adaptive to the signal/background ratio (R(S/B)) thresholding method for head and neck tumor delineation, and compare these GTV(PET) to reference GTV(CT) volumes in order to assess discrepancies. A cohort of 19 patients (39 lesions) with a histological diagnosis of head and neck cancer who would undergo definitive concurrent radiochemotherapy or radical radiotherapy with intensity-modulated radiotherapy technique (IMRT), were enrolled in this prospective study. Contouring on PET images was accomplished through standardized uptake value (SUV)-threshold definition. The threshold value was adapted to R(S/B). To determine the relationship between the threshold and the R(S/B), we performed a phantom study. A discrepancy index (DI) between both imaging modalities, overlap fraction (OF) and mismatch fraction (MF) were calculated for each lesion and imaging modality. The median DI value for lymph nodes was 2.67 and 1.76 for primary lesions. The OF values were larger for CT volumes than for PET volumes (p < 0.001), for both types of lesions. The MF values were smaller for CT volumes than for PET volumes (p < 0.001), for both types of lesions. The GTV(PET) coverage (OF(PET)) was strongly correlated with the lesion volume (GTV(CT)) for metastatic lymph nodes (Pearson correlation = 0.665; p < 0.01). For smaller lesions, despite the GTV volumes were relatively larger on PET than in CT contours, the coverage was poorer. Accordingly, the MF(PET/CT) was negatively correlated with the lesion volume for metastatic lymph nodes. The present study highlights the considerable challenges involved in using FDG PET imaging for the delineation of GTV in head and neck neoplasms. The methods that rely mainly on SUV(max) for thresholding, as the RS/B method, are very sensitive to partial volume effects and may provide unreliable results when applied on small lesions.