Improved PET lesion-detection performance using 2mm pixels

Positron emission tomography (PET) images are typically reconstructed with an in-plane pixel size of ~4mm for cancer imaging. The objective of this work was to evaluate the use of smaller pixels for the task of detecting focal warm lesions in a noisy structured background. Experimental phantom data from the Utah PET Lesion Detection Database was used, modeling whole-body oncologic FDG PET imaging of a ~92kg patient. The data comprised 24 scans over 4 days on a Biograph mCT TOF PET/CT scanner, with up to 23 lesions (diam. 6-16mm) distributed throughout the thorax, abdomen, and pelvis each day. Images were reconstructed with 2.036mm and 4.073mm pixels using ordered-subsets expectation-maximization (OSEM) both with and without point spread function (PSF) modeling and time-of-flight (TOF). Detection performance was assessed using the channelized non-prewhitened (CNPW) numerical observer with localization receiver operating characteristic (LROC) analysis. The observer was first used to optimize the number of iterations and smoothing filter for each case. Tumor localization performance and the area under the LROC curve (ALROC) were then analyzed as functions of pixel size. In all cases, the images with ~2mm pixels (ALROC = 0.59[OSEM], 0.60[PSF], 0.65[TOF] and 0.66[PSF+TOF]) provided higher detection performance than those with ~4mm pixels (ALROC = 0.56, 0.57, 0.61 and 0.63, respectively). The degree of improvement from using ~2mm pixels was larger than that from PSF modeling for these data, and provided roughly half the benefit of using TOF. Notably, all three effects cumulatively improved performance, and PSF+TOF reconstruction with 2mm pixels offered the best performance. This study suggests that a significant improvement in lesion-detection performance for general oncologic PET imaging can be attained by using smaller pixel sizes than current typical practice. The primary drawback is a ~4× increase in reconstruction time and data storage requirements.