Effects of New Bayesian Penalized Likelihood Reconstruction Algorithm on Visualization and Quantification of Upper Abdominal Malignant Tumors in Clinical FDG PET/CT Examinations

Mitsuaki Tatsumi; Fumihiko Soeda; Takashi Kamiya; Junpei Ueda; Daisuke Katayama; Keiko Matsunaga; Tadashi Watabe; Hiroki Kato; Noriyuki Tomiyama

doi:10.3389/fonc.2021.707023

Effects of New Bayesian Penalized Likelihood Reconstruction Algorithm on Visualization and Quantification of Upper Abdominal Malignant Tumors in Clinical FDG PET/CT Examinations

Front Oncol. 2021 Aug 16:11:707023. doi: 10.3389/fonc.2021.707023. eCollection 2021.

Authors

Mitsuaki Tatsumi^{1

2}, Fumihiko Soeda², Takashi Kamiya³, Junpei Ueda³, Daisuke Katayama², Keiko Matsunaga², Tadashi Watabe², Hiroki Kato², Noriyuki Tomiyama¹

Affiliations

¹ Department of Radiology, Osaka University Hospital, Suita, Japan.
² Department of Nuclear Medicine and Tracer Kinetics, Osaka University Graduate School of Medicine, Suita, Japan.
³ Department of Medical Technology, Osaka University Hospital, Suita, Japan.

Abstract

Purpose: This study evaluated the effects of new Bayesian penalized likelihood (BPL) reconstruction algorithm on visualization and quantification of upper abdominal malignant tumors in clinical FDG PET/CT examinations, comparing the results to those obtained by an ordered subset expectation maximization (OSEM) reconstruction algorithm. Metabolic tumor volume (MTV) and texture features (TFs), as well as SUV-related metrics, were evaluated to clarify the BPL effects on quantification.

Materials and methods: A total of 153 upper abdominal lesions (82 liver metastatic and 71 pancreatic cancers) were included in this study. FDG PET/CT images were acquired with a GE Discovery 710 scanner equipped with a time-of-flight system. Images were reconstructed using OSEM and BPL (beta 700) algorithms. In 58 lesions <1.5 cm in greatest diameter (small-lesion group), visual image quality of each lesion was evaluated using a four-point scale. SUVmax was obtained for quantitative metrics. Visual scores and SUVmax were compared between OSEM and BPL images. In 95 lesions >2.0 cm in greatest diameter (larger-lesion group), SUVmax, SUVpeak, MTV, and six TFs were compared between OSEM and BPL images. In addition to the size-based analyses, an increase of SUVmax with BPL was evaluated according to the original SUVmax in OSEM images.

Results: In the small-lesion group, both visual score and SUVmax were significantly higher in the BPL than OSEM images. The increase in visual score was observed in 20 (34%) of all 58 lesions. In the larger-lesion group, no statistical difference was observed in SUVmax, SUVpeak, or MTV between OSEM and BPL images. BPL increased high gray-level zone emphasis and decreased low gray-level zone emphasis among six TFs compared to OSEM with statistical significance. No statistical differences were observed in other TFs. SUVmax-based analysis demonstrated that BPL increased and decreased SUVmax in lesions with low (<5) and high (>10) SUVmax in original OSEM images, respectively.

Conclusion: This study demonstrated that BPL improved conspicuity of small or low-count upper abdominal malignant lesions in clinical FDG PET/CT examinations. Only two TFs represented significant differences between OSEM and BPL images of all quantitative metrics in larger lesions.

Keywords: Bayesian penalized likelihood reconstruction; PET; abdomen; malignant tumor; quantitative imaging.