Abstract
1002
Introduction: A major challenge of positron emission tomography (PET)/magnetic resonance imaging (MRI) is its lengthy examination owing to the whole-body PET scan and multiple MRI sequences. The Bayesian penalized likelihood (BPL) reconstruction (Q.Clear) permits full convergency without image degradation in a short emission time. An abbreviated MRI (abb-MRI), a shortened version of the standard MRI (std-MRI) with a reduced number of sequences, enables quick whole-body MRI scan. This study aimed to assess the optimal beta (β) value of time-of-flight (TOF) BPL in a short emission scan duration and evaluate the ability of quick whole-body PET/MRI scan for detecting and differentiating lesions in patients with malignancy using BPL and abb-MRI.
Methods: Fifty-three whole-body 18F-fluorodeoxyglucose PET/MRI scans with list mode acquisition, performed for consecutive patients with pathologically confirmed malignancy, were retrospectively analyzed. The optimized β values for BPL with short emission time (reconstructed in 1.5 and 1.0 min per bed position) were qualitatively and quantitatively determined by statistical comparisons with TOF ordered-subset expectation maximization (OSEM) reconstruction in standard emission time (2.5 min). To assess the ability to detect and differentiate lesions, two readers trained in radiology and nuclear medicine evaluated the abnormal lesions in terms of detectability and differentiation between benign and malignant lesions using 5-point visual scores. Each evaluated PET/MRI scan involved a short scan (BPL and abb-MRI) and a standard scan (OSEM and std-MRI) and was separately analyzed in four different regions (the head and neck, chest, abdomen, and pelvis). The reference standards were acquired using histopathological results and/or follow-up radiological findings during the 6-month follow-up. Statistical comparison was performed using receiver operating characteristic (ROC) analysis followed by McNemar’s test. Inter-reader agreement was tested using Cohen’s kappa coefficient.
Results: BPL with β600 in 1.5-min scan (BPL1.5) and that with β700 in 1.0-min scan (BPL1.0) showed no significant qualitative and quantitative difference with standard OSEM in 2.5-min scan (OSEM2.5). No significant difference was observed in the area under the ROC curve (AUC) for the detection ability of BPL1.5/abb-MRI, BPL1.0/abb-MRI, and OSEM2.5/abb-MRI with OSEM2.5/std-MRI (Reader 1, p=0.146, p=0.146, and p=0.150; Reader 2, p=0.146, p=0.146, and p=0.146, respectively). Moreover, the 95% confidence intervals of the AUC difference between those were more than the fixed margin of -6% for the non-inferiority test. An almost perfect agreement (weighted kappa=0.945) was observed in the inter-rater assessments for detection ability. Similarly, no significant difference was observed in the AUC for the differentiation ability of BPL1.5/abb-MRI, BPL1.0/abb-MRI, and OSEM2.5/abb-MRI with OSEM2.5/std-MRI (Reader 1, p=0.330, p=0.330, and p=0.359; Reader 2, p=0.147, p=0.147, and p=0.144, respectively). In addition, the 95% confidence intervals of the AUC difference between those were more than the fixed margin of -6% for the non-inferiority test. An almost perfect agreement (weighted kappa=0.925) was observed in the inter-rater assessments for the differentiation ability. Conclusion: A combination of TOF-BPL with optimal β value and abb-MRI enables a quick whole-body PET/MRI examination in <1.5 min per bed position, with equivalent ability for the detection and differentiation of lesions to the conventional TOF-OSEM in a 2.5-min scan duration with std-MRI.