Automated biological target volume delineation for radiotherapy treatment planning using FDG-PET/CT

Background: This study compared manually delineated gross tumour volume (GTV) and automatically generated biological tumour volume (BTV) based on fluoro-deoxy-glucose (FDG) positron emission tomography (PET)/CT to assess the robustness of predefined PET algorithms for radiotherapy (RT) planning in routine clinical practice. Methods: RT-planning data from 20 consecutive patients (lung-(40\%), oesophageal-(25\%), gynaecological-(25\%) and colorectal (10\%) cancer) who had undergone FDG-PET/CT planning between 08/2010 and 09/2011 were retrospectively analysed, five of them underwent neoadjuvant chemotherapy before radiotherapy. In addition to manual GTV contouring, automated segmentation algorithms were applied-among these 38\%, 42\%, 47\% and 50\% SUVmax as well as the PERCIST total lesion glycolysis (TLG) algorithm. Different ratios were calculated to assess the overlap of GTV and BTV including the conformity index and the ratio GTV included within the BTV. Results: Median age of the patients was 66 years and median tumour SUVmax 9.2. Median size of the GTVs defined by the radiation oncologist was 43.7 ml. Median conformity indices were between 30.0-37.8\%. The highest amount of BTV within GTV was seen with the 38\% SUVmax algorithm (49.0\%), the lowest with 50\% SUVmax (36.0\%). Best agreement was obtained for oesophageal cancer patients with a conformity index of 56.4\% and BTV within GTV ratio of 71.1\%. Conclusions: At present there is only low concordance between manually derived GTVs and automatically segmented FDG-PET/CT based BTVs indicating the need for further research in order to achieve higher volumetric conformity and therefore to get access to the full potential of FDG-PET/CT for optimization of radiotherapy planning