Efficacy Study of Deep Progressive Reconstruction Algorithm in Enhancing 18F-FDG PET Image Quality for Breast Cancer Patients with Different Body Mass Index

Objective To evaluate the efficacy of Deep Progressive Reconstruction (DPR) versus Ordered Subset Expectation Maximization (OSEM) algorithms in enhancing ^18F-FDG PET image quality across different body mass index (BMI) strata in patients with breast cancer. Methods This retrospective study included patients with breast cancer who underwent diagnostic ^18F-FDG PET/CT at the China-Japan Union Hospital of Jilin University between June and December 2023. Whole-body PET/CT was performed using the United Imaging uMI 780 system, with images reconstructed using three algorithms: DPR, OSEM2 (two iterations), and OSEM3 (three iterations). Patients were stratified into four BMI subgroups: underweight, normal weight, overweight, and obese. Two board-certified nuclear radiologists independently and blindly assessed image quality, noise level, and lesion conspicuity using a 5-point Likert scale. Quantitative metrics included maximal lesion diameter, maximum standardized uptake value (SUVmax), peak SUV (SUVpeak), liver signal-to-noise ratio (LSNR), tumor-to-background ratio (T/N), and liver SUV standard deviation (SUVSD). Statistical analyses were performed with one-way analysis of variance (ANOVA) and Mann–Whitney U tests to compare the three reconstruction methods across BMI subgroups and lesion sizes. Results A total of 118 female patients with 188 measurable lesions were analyzed. DPR-reconstructed PET images achieved significantly higher quality scores and lower noise levels than OSEM2 and OSEM3 (all P < 0.05). Rising BMI significantly worsened image quality and noise scores in OSEM2/OSEM3 (P < 0.05), with OSEM3 in obese patients failing to meet diagnostic criteria (quality score: 2.6; noise score: 2.7). DPR reconstruction yielded significantly higher SUVmax, SUVpeak, LSNR, and T/N values, while producing markedly lower SUVSD compared to OSEM methods (all P < 0.05). Compared with OSEM3, DPR conferred greater improvements in SUVmax, SUVpeak, and T/N in low-weight patients (P = 0.011, 0.048, and 0.014, respectively), and larger gains in SUVSD and LSNR in high-weight patients (P = 0.003 and < 0.001, respectively). Inverse correlations were observed between lesion size and DPR improvements versus OSEM2 for SUVmax (r=–0.298), SUVpeak (r=–0.243), and T/N (r=–0.326) (all P < 0.01), and versus OSEM3 for SUVmax (r=–0.213, P < 0.01). When compared with OSEM2, DPR provided greater benefits for sub-2 cm lesions (+ 12% SUVmax, + 6% SUVpeak, + 17% T/N) than for larger lesions (+ 5%, + 1%, + 6%; all P < 0.001). Conclusion DPR significantly improved ^18F-FDG PET image quality, noise suppression, and lesion conspicuity compared with OSEM. These advantages were most pronounced in patients with higher BMI and in small lesions, highlighting the potential of DPR to enhance diagnostic performance in challenging patient subgroups.