The value of Bayesian penalized likelihood reconstruction for improving lesion conspicuity of malignant lung tumors on 18F-FDG PET/CT: comparison with ordered subset expectation maximization reconstruction incorporating time-of-flight model and point spread function correction

概要

Objective
　To evaluate the usefulness of the Bayesian penalized likelihood (BPL) reconstruction method for improvement of lesion conspicuity of malignant lung tumors, compared with the ordered subset expectation maximization (OSEM) reconstruction incorporating both time-of-flight (TOF) model and point-spread- function (PSF) algorithm on 18F-fluoro-2-deoxy-D-glucose (FDG) positron emission tomography computed tomography (PET/CT).

Methods
　Twenty-nine patients with primary or metastatic lung cancers who underwent 18F-FDG PET/CT were retrospectively studied. PET images were reconstructed with six different algorithms -l)TOF OSEM with PSF, 2)TOF OSEM without PSF, 3)BPL with β-value of 200, 4)BPL with β-value of 400, 5)BPL with β-value of 600, 6)BPL with β-value of 800. The signal-to-noise ratio (SNR) was determined in normal liver parenchyma. Lung lesion conspicuity of 50 malignant lung nodules was visually graded on a 4-point scale(0, no visible；1,poor； 2, good； 3, excellent conspicuity) by 2 independent observers. These 2 observes were also independently asked to choose the most preferred reconstruction for detecting the lung lesions on a per-patient level. The maximum standardized uptake value (SUVmax) was measured, in each of the 50 lung lesions.

Results
　SNR on the images reconstructed by BPL with β-value of 600 and 800(17.8 ± 3.7 and 22.5 ± 4.6, respectively) was significantly higher than that by OSEM + TOF + PSF (15.0 ± 3.4, p < 0.0001). BPL with β-value of 600 was chosen most frequently as the preferred reconstruction algorithm for lung lesion assessment by both observers. The lesion conspicuity of the lung nodules < 10 mm in diameter was significantly improved by BPL with β-value of 600, compared to OSEM + TOF + PSF(2.2 土 0.8 vs 1.6 土 0.9, p < 0.0001),while the conspicuity score of the lesions < 10 mm in diameter was not significantly different between BPL with β-value of 600 and OSEM + TOF + PSF. The mean SUVmax of the lung nodules with < 10 mm on the images reconstructed by BPL with β-value of 600 was greater than that by OSEM + TOF + PSF(4.2 土 3.5 vs 3.4 ± 3.1,p = 0.001). Regarding the lung lesions < 10 mm in diameter, BPL with β-value of 600 did not provide increased SUVmax, compared to OSEM + TOF + PSF reconstruction.

Conclusion
　BPL reconstruction algorithm, particularly with β-value of 600, significantly improves lesion conspicuity of malignant lung nodules with < 10 mm in diameter, compared to OSEM+TOF+PSF reconstruction. BPL can be of particularly added value in assessing small lung malignant lesions on 18F_FDG PET/CT images.