Bayesian penalized-likelihood iterative reconstruction of PET/CT for evaluation of lung nodules.

2735 Objectives The goal of our research was to compare the recently FDA approved Bayesian penalized-likelihood (BPL) iterative reconstruction of PET/CT to current standard forms of iterative reconstruction. Specifically, we asked if BPL improves evaluation of lung nodules. Methods FDG PET/CT scans for the indication of “lung nodule” from 8/5/14 to 12/31/14 were retrospectively reviewed. Scans of patients previously treated for cancer were excluded. All patients reviewed were scanned on a GE Discovery PET/CT 710 camera per standard clinical protocol. All PET scans were reconstructed using 3D OSEM, Time of Flight (TOF), and BPL. For each lung nodule SUVmax and an adjacent background lung activity was measured. Additional measurements were made in the blood pool and liver. When SUVmax of a nodule was greater than blood pool, it was called “positive”. Images were interpreted on an OsiriX 64 bit workstation and a Nuclear Radiologist confirmed all measurements. Statistical analysis was performed with JMP software. Results 34 lung nodules in 20 patients were reviewed (mean age 73 +/- 7.5 yrs; glucose 107 +/- 25 mg/dL; BMI 26.5 +/- 5.0). The average nodule diameter was 20.7 +/- 17.5 mm. 23 (68%) of nodules were “positive” on 3D, 27 (79%) on TOF, and 29 (85%) on BPL reconstructed images. Nodules below 5mm were not "positive" on any reconstruction. The mean SUVmax was 6.7 +/- 5.1 for 3D, 7.0 +/- 5.0 for TOF (11.3 +/- 16.5% increase over 3D, p Conclusions FDG PET/CT reconstructed with BPL resulted in significantly increased SUVmax and SNR of lung nodules. This resulted in higher numbers of nodules and smaller nodules being identified as “positive”.