TY - JOUR T1 - Lesion detectability in long axial field of view TOF PET scanners JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 107 LP - 107 VL - 60 IS - supplement 1 AU - Varsha Viswanath AU - Margaret Daube-Witherspoon AU - Joel Karp AU - Suleman Surti Y1 - 2019/05/01 UR - http://jnm.snmjournals.org/content/60/supplement_1/107.abstract N2 - 107Objectives: Lesion detectability is an important clinical metric that reflects a clinician’s ability to detect small, low contrast lesions. The improved sensitivity of long axial field of view (AFOV) PET should translate to reduced statistical noise and, therefore, improved lesion detectability. Our group is currently building the PennPET Explorer, which is based on digital PET detectors and currently has three rings, each 23-cm in length. To understand the performance of the 70-cm long AFOV scanner, we performed GATE simulations and ran basic NEMA tests to compare with measured results. These GATE simulations showed a sensitivity of 90.5 kcps/MBq for the 70-cm scanner and a per-organ sensitivity gain of 2.7 at the center of the AFOV when compared to a 23-cm ring. The goal of this work was to quantify detectability of anthropomorphic distributions on long AFOV systems as a function of collected counts (i.e. scan time), lesion size, and lesion contrast. Methods: Previous studies of lesion detectability using a simple activity distribution for a single ring vs. 70-cm AFOV showed improvement in the area under the LROC curve (ALROC) that reflected the per-organ sensitivity gain in the center of the scanner. To quantify lesion detectability on a long AFOV system, we used the 4D extended cardiac-torso (XCAT) phantom to simulate an adult female injected with 10 mCi of FDG and imaged 1-hour post injection. The phantom was imaged from skull base to pelvis on the 70-cm PET scanner for 3 min with 7-mm liver lesions at 2:1, 3:1, and 4:1 local contrast. Data were subsampled to scan times as low as 15 sec and reconstructed using blob-based list-mode ordered subset expectation maximization (4 iterations x 25 subsets) into 2x2x2 mm3 voxels. On the PennPET Explorer we imaged a normal volunteer (positioned apex to mid-belly) injected with 14.9 mCi of FDG and imaged 107 min after injection for 20 min. The data were subsampled into six replicates with scan times of 3 mins or less. Separately acquired 6-mm and 10-mm spheres were combined with the human data to emulate lesions in the subject’s liver and lung at several contrast levels. Lesion localization and detectability were numerically estimated for the measured and simulated images using a generalized scan statistics methodology and quantified with the ALROC metric. Results: Generalized scan statistics methodology was successfully implemented in both the liver and lung, a non-uniform background, and was used to characterize detectability of small lesions as a function of scan time. CRCs of the small lesions are consistent with previously measured and simulated values, indicating consistency of the simulated and measured data. Detectability, as measured by ALROC, was high (≥ 0.8) for scan times as low as 60 sec for both simulated and measured data in the larger spheres and at higher contrasts. However, the ALROC for the 6-mm spheres in the liver was lower than predicted by simulation, and further investigation is planned. When comparing similar lesions in the simulation and measurement, ALROC values were consistent after scaling for differences in collected counts. These results imply that we can expect detectability to scale based on differences in sensitivity. Conclusions: This work has studied lesion detectability in anthropomorphic phantoms and human subjects on a long AFOV scanner and shown stable detectability for scan times between 60 s and 3 min. However, the results show that detection of small (6-7 mm) lesions is challenging, even with the higher sensitivity achieved with a long AFOV. Because detectability is affected by collected counts, which can vary as function of the axial location of lesions, we will study the effect of axial location on ALROC. The relationship between ALROC and axial location will predict detectability for single bed-position imaging on long AFOV scanners, key for both static and dynamic imaging, and guide ongoing expansion of the PennPET Explorer past 100 cm. ER -