Clinical evaluation of regularized OSEM reconstruction algorithm in pediatric total-body ¹⁸F-FDG PET/CT

Objectives: To evaluate the effects of regularized ordered-subsets expectation maximization reconstruction algorithm (R-OSEM) on total-body ¹⁸F- FDG PET/CT imaging for pediatric patients with cancer.

Methods: A total of 12 pediatric patients (age 4.9±1.9 years, weight 17.94±5.75 kg) with cancer have been enrolled. Using 194-cm-long axial field-of-view PET/CT scanner (uEXPLORER, United Imaging Healthcare), all patients underwent 10-min ¹⁸F-FDG PET examinations with the administered dose of 3.7 MBq/kg. Various total-body PET images with reduced count density have been reconstructed by truncating the 3D list-mode PET data (5 min, 3 min, 1 min, 40 s, and 20 s), where the applied reconstruction algorithms are the R-OSEM with time-of-flight correction (R-OSEM-TOF) and the time-of-flight ordered-subsets expectation maximization with point-spread function (OSEM-TOF-PSF). In order to determinate the optimal regularization strength (β) for different acquisition times, R-OSEM data sets with a β value of 0.1, 0.28, 0.5 and 0.9 were involved in this study. For subjective evaluation, the average of standard uptake value (SUV_mean) and standard variation (SD) at liver, the maximum of standard uptake value (SUV_max) at lesions (lesion size 8.7±2.11 mm), as well as the lesion-to-background ratio (LBR) have been recorded and compared with OSEM data.

Results: Exploiting R-OSEM-TOF algorithm, the improvement of LBR compared with OSEM-TOF-PSF has been observed in total-body ¹⁸F-FDG PET images of pediatric patients, revealing a great potential in small lesion detectability. Indeed, R-OSEM-TOF increased the SUV_max of lesions and maintained the SUV_mean at liver when the acquisition time above 40 s, resulting in an optimized LBR. Despite noise effect enhanced with reducing acquisition time, the use of ROSEM-TOF declined the SD at liver, especially for PET imaging with low statistic counts. Such impact leads to the ROSEM-TOF-reconstructed 20-s PET snapshot (β = 0.28) with image quality similar to OSEM-TOF-PSF PET of 1 min through comparing SD, SUV_max as well as LBR (no significant difference by paired t-test with Bonferroni correction).

Conclusions: Compared with clinically widespread OSEM-TOF-PSF, R-OSEM-TOF illustrates a particular potential in the improvement of small lesion detectability and the decreasing of acquisition time for PET examination in pediatric patients. Combining both R-OSEM-TOF and uEXPLORER, one might achieve a PET image meeting with clinical requirements, and simultaneously, decrease dramatically either the administered dose for minimizing the risk derived from the ionization radiation or the acquisition time for fast PET imaging without sedation in pediatric patients.

• Download figure
• Open in new tab
• Download powerpoint