Appropriately regularized OSEM can improve the reconstructed PET images of data with low count statistics

With the increasing number of patients undergoing positron emission tomography (PET) scans and the fact that multiple whole body acquisitions are performed during therapy monitoring, the reduction of scan time as well as of the injected radioactive dose are important issues. However, short scan time and reduction of the injected radiation dose result in low count statistics, which significantly affects the quality of the reconstructed images and accurate diagnosis. The aim of this study was to explore the effect of low count statistics on ordered subset expectation maximization regularized with median root prior (OS-MRP-OSL) reconstructed images. By optimizing OS-MRP-OSL we determine whether a satisfactory handling of the noise properties and bias can be achieved compared to post-filtered ordered subset expectation maximization (OSEM), which will lead to improved image quality in simulations with more noise. We used realistic simulated PET data of a thorax with lesions corresponding to tumors with different intensities. OS-MRP-OSL provided reduced noise from post-filtered OSEM, without having the negative effect of blurring. On the other hand, bias presented no significant difference. This work is relevant to future PET reconstruction of clinical images and PET-magnetic resonance investigations where the reduced injected dose will allow imaging a larger cohort of humans.