PD-0293 REPRODUCIBILITY AND VOLUME SEGMENTATION METHODS ON F-MISO AND FLT PET-CT IMAGES IN PATIENTS WITH LUNG CANCER

Purpose/Objective: Our objectives were (i) to investigate whether different Nuclear Medicine centres would rate similarly 18 Fmisonidasole (F-miso hypoxia) or 18fluorothymidine (FLT proliferation) uptake and (ii) to compare different segmentation methods, using F-miso and FLT. Materials and Methods: Twenty PET-CT (10 F-miso and 10 FLT) were performed before and during curative-intent radiotherapy in 5 patients with NSCLC. The images were electronically sent to 18 centres of Nuclear Medicine connected through the French Society of Nuclear Medicine network (SFMNnet). In a first phase, each centre had to rank the F-miso and FLT uptake intensity on a 5-level scale, and secondly into a 2-level scale (0=normal, 1=pathologic). The concordance between readers was quantified by weighted kappa tests (K). The presence or absence of F-miso and FLT uptake defined by experts was used as gold standard. In a second phase, potential target F-miso and FLT volumes were delineated with 22 different methods of segmentation: a threshold of 40% and 50% of the maximum standardized uptake value (SUVmax), a threshold fixed at 1.2 and 1.4 of the SUVmax, and 3 methods using an adaptive thresholding related to the SUVmax or SUVmean of a reference tissue (muscle, aorta and lung). Results: As for uptake intensity, the K values between the centres were 0.44 for F-miso and 0.43 for FLT with a 5-level scale, and 0.77 for F-miso and FLT with a 2-level scale. The various segmentation methods yielded significantly different volumes for F-miso and FLT (p max of aortic uptake (sensitivity: 0.87 and 0.71, specificity: 0.72 and 0.96 respectively). For FLT, 6 methods (fixed threshold at 1.4, adaptive threshold based on 1.5 or 1.6 of aortic SUVmax, and adaptive threshold on based 1.3, 1.5 or 1.6 muscle SUVmax) gave the best agreement. Conclusions: The 18 centres were able to reproducibly grade F-miso and FLT uptake on a 2-level (positive/negative) scale. Our data support a fixed threshold (1.4 of aortic SUVmax) to delineate increased F-miso uptake and an adaptive threshold using the aorta and muscle SUVmax background for FLT.