Quantitative evaluation of atlas-based attenuation correction for brain PET in an integrated time-of-flight PET/MR imaging system

Purpose: To assess the patient-dependent accuracy of atlas-based attenuation correction (ATAC) for brain positron emission tomography (PET) in an integrated time-of-flight (TOF) PET/magnetic resonance (MR) imaging system. Materials and Methods: Thirty recruited patients provided informed consent in this institutional review board-approved study. All patients underwent whole-body fluorodeoxyglucose PET/computed tomography (CT) followed by TOF PET/MR imaging. With use of TOF PET data, PET images were reconstructed with four different attenuation correction (AC) methods: PET with patient CT-based AC (CTAC), PET with ATAC (air and bone from an atlas), PET with ATAC_patientBone (air and tissue from the atlas with patient bone), and PET with ATAC_boneless (air and tissue from the atlas without bone). For quantitative evaluation, PET mean activity concentration values were measured in 14 1-mL volumes of interest (VOIs) distributed throughout the brain and statistical significance was tested with a paired t test. Results: The mean overall difference (6standard deviation) of PET with ATAC compared with PET with CTAC was 20.69 kBq/mL ± 0.60 (24.0% ± 3.2) (P < .001). The results were patient dependent (range, 29.3% to 0.57%) and VOI dependent (range, 25.9 to 22.2). In addition, when bone was not included for AC, the overall difference of PET with ATAC_boneless (29.4% ± 3.7) was significantly worse than that of PET with ATAC (24.0% ± 3.2) (P < .001). Finally, when patient bone was used for AC instead of atlas bone, the overall difference of PET with ATAC_patient-Bone (21.5% ± 1.5) improved over that of PET with ATAC (24.0% ± 3.2) (P < .001). Conclusion: ATAC in PET/MR imaging achieves similar quantification accuracy to that from CTAC by means of atlas-based bone compensation. However, patient-specific anatomic differences from the atlas caused bone attenuation differences and misclassified sinuses, which result in patient-dependent performance variation of ATAC.