Improvement of image quality for small lesion sizes in

Objectives: We investigated image quality and standardized uptake values (SUVs) for different lesion sizes using clinical data generated by Methods: We evaluated the effect of point-spread function (PSF) modeling and Gaussian filtering (Gau) and determined the optimal reconstruction conditions. We compared the signal-to-noise ratio (SNR), contrast, %coefficient of variation (%CV), SUV, and Likert scale score between ordered-subset expectation maximization (OSEM) time-of-flight (TOF) and OSEM+TOF+PSF in phantom and clinical studies. The conventional image was generated with OSEM+TOF_Gau 6 mm. The National Electrical Manufacturers Association body phantom with 10-mm hot sphere data was acquired for 5 min. Twenty-six patients (40 lesions, ranging from 3.7 to 63.0 mm) were examined using prone breast PET/CT with a breast positioner for breast cancer staging. PET data were acquired 125±9.7 min after intravenous injection of 220±16.1 MBq at 5 min/bed. Results: In the phantom study, a high SNR was obtained from a 3- to 5-mm Gaussian filter for OSEM+TOF+PSF. The contrast obtained with OSEM+TOF without Gaussian filtering was superior to that obtained with OSEM+TOF+PSF_Gau 4 mm. In the clinical study, the image quality depended on lesion size. The average SNR was significantly higher at 40.8% for lesions >20 mm with OSEM+TOF_Gau 6 mm than with OSEM+TOF without Gaussian filtering. The average contrast for lesions ≤10 mm was significantly higher by 42.0% with OSEM+TOF without Gaussian filtering than with OSEM+TOF_Gau 6 mm. The average SUV Conclusion: OSEM+TOF without Gaussian filtering provided good contrast and quantitative value for small lesions.