Data-Driven Motion Correction of 11C-Methionine PET Images on a Cohort of Pediatric Patients With Brain Tumor in a PET/MRI Study

A data-driven head Motion Correction (MoCo) method was tested on 11C-Methionine brain PET images in a cohort of 44 pediatric patients with brain tumors referred to a PET/MR study. Its impact was investigated both qualitatively and quantitatively. For each patient, PET images were reconstructed offline both with (PETddMoCo) and without (PETnoMoCo) MoCo algorithm. An expert Nuclear Medicine physician qualitatively evaluated PET images, and segmented PET positive lesions in both datasets, extracting the following PET parameters: maximum and mean Standardized Uptake Value (SUVmax and SUVmean, respectively), and Metabolic Tumor Volume (MTV). PET parameters before and after MoCo were compared and their absolute percentage difference was calculated (Δ%). Contrast-to-Noise (CNR) and its difference (Δ) before and after MoCo were calculated. Thirty-one patients had a “low” level of motion, 8 patients “medium” and 5 patients “high”. Twenty-one patients out of forty-four had positive 11C-Methionine uptake (26 lesions). Qualitatively, no difference was evident in negative patients, while two PET positive lesions could be better defined after motion correction. Quantitatively, CNR increased significantly after motion correction for “medium+high” lesions, while none of the PET parameters showed significant difference. Increasing the sample of patients might confirm these results.