A contrast-oriented algorithm for FDG-PET-based delineation of tumour volumes for the radiotherapy of lung cancer: derivation from phantom measurements and validation in patient data

PurposeAn easily applicable algorithm for the FDG-PET-based delineation of tumour volumes for the radiotherapy of lung cancer was developed by phantom measurements and validated in patient data.MethodsPET scans were performed (ECAT-ART tomograph) on two cylindrical phantoms (phan1, phan2) containing glass spheres of different volumes (7.4–258 ml) which were filled with identical FDG concentrations. Gradually increasing the activity of the fillable background, signal-to-background ratios from 33:1 to 2.5:1 were realised. The mean standardised uptake value (SUV) of the region-of-interest (ROI) surrounded by a 70% isocontour (mSUV70) was used to represent the FDG accumulation of each sphere (or tumour). Image contrast was defined as:$$C = {{\left( {{\text{mSUV}}_{{\text{70}}} - {\text{BG}}} \right)} \mathord{\left/ {\vphantom {{\left( {{\text{mSUV}}_{{\text{70}}} - {\text{BG}}} \right)} {{\text{BG}}}}} \right. \kern-\nulldelimiterspace} {{\text{BG}}}}$$ where BG is the mean background − SUV. For the spheres of phan1, the threshold SUVs (TS) best matching the known sphere volumes were determined. A regression function representing the relationship between TS/(mSUV70 − BG) and C was calculated and used for delineation of the spheres in phan2 and the gross tumour volumes (GTVs) of eight primary lung tumours. These GTVs were compared to those defined using CT.ResultsThe relationship between TS/(mSUV70 − BG) and C is best described by an inverse regression function which can be converted to the linear relationship $${\text{TS}} = a \times {\text{mSUV}}_{70} + b \times {\text{BG}}$$. Using this algorithm, the volumes delineated in phan2 differed by only −0.4 to +0.7 mm in radius from the true ones, whilst the PET-GTVs differed by only −0.7 to +1.2 mm compared with the values determined by CT.ConclusionBy the contrast-oriented algorithm presented in this study, a PET-based delineation of GTVs for primary tumours of lung cancer patients is feasible.