Diagnostic performance of rapid whole-body FDG PET/MRI in oncology patients: Assessment of optimal beta-value for block sequential regularized expectation maximization (BSREM) reconstruction and combination of whole-body MR sequences

Purpose: The prolonged examination time for PET/MRI, due to the requirement for whole-body and regional scans, is a major drawback. A Block sequential regularized expectation maximization (BSREM) reconstruction (Q.Clear) controls image noise and enables a short emission time without image degradation, to yield a fully convergent PET image. The primary purpose of the study was to assess the optimized beta (β) value of time-of-flight (TOF) BSREM, for a rapid emission scan. The secondary purpose was to evaluate the diagnostic performance of the BSREM with a determined β value and selected MR sequences in comparison with a conventional TOF ordered-subsets expectation maximization (OSEM) reconstruction with standard emission time and full MR sequences. MATERIALS AND METHODS: Fifty-four patients with pathologically confirmed malignancy, who consecutively underwent whole-body F-18 FDG PET/MRI, were retrospectively evaluated. Three different emission scan durations (2.5 min, 1.5 min, and 1.0 min) and four different β values (500, 600, 700, and 800) were applied to the list mode reconstructions of PET, with point-spread function recovery. To assess the optimal β value, BSREM in a short scan duration (1.5 min and 1.0 min) were statistically compared with OSEM (16 subsets, 2 iterations, and 4 mm Gaussian filter) in a standard scan duration (2.5 min). Two nuclear medicine physicians evaluated the image quality against 5-point visual scores for liver noise level, SUV of the lesions, signal to background ratio (SBR), and signal to noise ratio (SNR). To determine the diagnostic performance for BSREM with determined β value and selected MR sequences acquirable during emission scan, the images obtained from BSREM and OSEM in different scan durations were visually evaluated by two readers with 5-point visual scores regarding differentiation of each lesion between benign and malignant. The reference standards were acquired by histopathological results and/or follow-up radiological examinations, during the 6 month follow-up period.

Results: A total of 257 lesions in 54 patients were evaluated for the primary purpose. According to the Wilcoxon’s signed rank test and the Bland and Altman plot, no significant differences in visual scores for liver noise level, lesion SUVmax, and lesion SBR, were observed between BSREM with β value between 500 and 600 in 1.5 min scan duration, BSREM with β value between 600 and 700 in 1.0 min scan duration, and OSEM in 2.5 min scan duration. To avoid a clustering bias for further statistical analysis, 173 lesions in 54 patients were randomly selected. The receiver operating characteristics curve showed that the area under the curve (AUC) for BSREM with β600 in 1.5 min scan duration, and β700 in 1.0 min scan duration was significantly larger than the AUC for OSEM in 1.5 min scan duration for experienced reader (p=0.0204 and p=0.0265, respectively) and OSEM in 1.0 min scan duration for both, experienced (p=0.0030 and p=0.0031, respectively) and novice reader (p=0.0372 and p=0.0372, respectively). No significant difference was observed for AUC between BSREM in 1.5 min and 1.0 min scan durations and OSEM in 2.5min (experienced, p=0.7758 and p=0.6645; novice, p=0.7745 and p=0.7745, respectively) scan duration. Moreover, the 95% confidence intervals of the difference (experienced, -0.0117-0.0157 and -0.0109-0.0171; novice, -0.0189-0.0253 and -0.0189-0.0253, respectively) were more than the fixed-margin of -2% for the noninferiority test.

Conclusions: TOF-BSREM with optimal β value and selected MR sequences enables a rapid whole-body PET/MRI examination in less than 1.5 minutes per bed position. No image degeneration occurs, and the diagnostic performance is equivalent to the conventional TOF-OSEM in 2.5 min.