TY - JOUR T1 - <strong>Design and construction of a heterogeneous FDG uptake phantom for tumor volume delineation as an insert to the NEMA image quality phantom</strong> JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 388 LP - 388 VL - 60 IS - supplement 1 AU - Ana Poma AU - Franco Passian AU - Virginia Venier AU - Humberto Romano AU - Pablo Goldaraz AU - Javier Promet AU - Tania Telford AU - Camilo Garcia Y1 - 2019/05/01 UR - http://jnm.snmjournals.org/content/60/supplement_1/388.abstract N2 - 388Introduction: Metabolic PET images are being used more and more often for tumor delineation in radiotherapy planning. Tumor volume delineation based on automatic, or semi-automatic methods, results in in lower inter- and intra-observer variability. Publications which describe PET segmentation methods are mainly based on phantom images and the methods most frequently used include: fixed/adaptive thresholding (based on SUVs), gradient tumor volume delineation, deep learning algorithms and supervised segmentation methods. These methods were developed using either uniform uptake phantoms, Monte Carlo simulated phantoms with differing structural and noise-simulation complexity and/or phantoms which contained simple geometrical structures. Objectives: We describe the design and construction of three inserts for the NEMA image quality phantom, which allow us to carry out homogeneous and heterogeneous volume delineation. We also describe the subsequent acquisition of PET/CT images using these inserts. Methods: Three inserts were designed: an empty one, another one in the shape of a female pelvis of reduced size and a third one containing different sections, similar in appearance to a bicycle chain. They were all designed with elliptical bases and have axial dimensions of 14.5 x 11.5 cm. This allows them to fit in the larger NEMA image quality phantom, whose top insert has a diameter of 15 cm. The inserts are 3, 3, and 4.8 cm tall respectively. They were designed to be stacked one on top of each other over the section of the NEMA phantom usually reserved for the lung insert. The empty insert allows materials with different water absorption indices to be placed inside it. This results in a heterogeneous, non-controlled uptake pattern, similar to those produced by tumors. The second insert (the female pelvis) was designed to evaluate the performance of segmentation methods used when delineating cervical tumor volumes near the area of high bladder uptake. The third insert contains two adjacent chambers, each of which contains successively smaller internal concentric and non-concentric chambers. The chambers may all be filled with different radioactive concentrationsor left empty, giving extra versatility to the phantom. The inserts were constructed by 3D printing of transparent polylactic acid. Their walls are 2.5 mm thick, with 0.6mm thick lids and bases. Locking cylinders were also printed. These were 3 mm thick and 5 cm tall (the highest and lowest cylinders) and 5 mm tall (the middle cylinders). Results: The inserts were used to produce a number of images with different patterns. These ranged from non-realistic, sphere-like distributions, to uncontrollable, heterogeneous patterns, similar to those found in patients. The “bicycle chain” insert was used to analyze different activity gradients, which allowed us to model necrosis-like situations in tumors. We were also able to analyze images with different tumor to background ratios. This is important as low ratios are known to degrade the quality of volume contours. Conclusions: The designed inserts may be used to simulate different distributions of F-18 uptake. This ranges from uniform and controllable uptake patterns, to those which mimic the heterogeneous distributions seen in pathologies. The information obtained from these analyses may then be used for carrying out tumor delineation in a radiotherapy setting. ER -