Scholarlink

Background <p>To investigate the impact of deep progressive learning reconstruction (DPR) on small lesion detection and image quality compared to ordered subset expectation maximum (OSEM) and regularized OSEM (ROSEM) in <sup>18</sup>F-FDG PET/CT imaging.</p> Methods <p>The NEMA phantom was filled with <sup>18</sup>F-FDG solution, with a hot sphere-to-background ratio of 4:1. Twenty-six patients with <sup>18</sup>F-FDG-avid lung lesions (diameter &lt; 2.0&#xa0;cm) were enrolled in the study. The PET images were reconstructed by seven groups: routine OSEM, ROSEM with a penalization factor of 0.8 (ROSEM), and DPR reconstructions with five different filter strength factors ranging from smooth to sharp:1–5 (DPR1, DPR2, DPR3, DPR4, and DPR5). The contrast recovery (CR), background variability (BV), contrast-to-noise ratio (CNR), and radioactivity concentration ratio (RCR) were measured in the phantom study. The maximum standardized uptake value (SUV<sub>max</sub>), target-to-background ratio (TBR), CNR, the volume of the lesions, and the coefficient of variation (COV) of the liver were calculated and compared between these methods in the patient study. Two radiologists evaluated the image quality using a five-point Likert scale.</p> Results <p>In the phantom study, the DPR2 to DPR5 and ROSEM groups achieved higher CR, CNR, RCR, and lower BV than the OSEM group. For the smallest 10-mm hot sphere, DPR3 achieved a CNR of 24.46, compared to 22.25 for ROSEM and 10.39 for OSEM. In patient studies, liver background noise (COV) was significantly lower in DPR1-DPR4 compared to OSEM (all <i>p</i> &lt; 0.05), with no significant difference between the DPR3 and ROSEM (<i>p</i> = 0.65). Lesion quantification metrics-SUV<sub>max</sub>, TBR, and CNR were significantly higher in DPR2-DPR4 than in OSEM (all <i>p</i> &lt; 0.05). Notably, DPR4 and DPR5 provided equivalent lesion SUV<sub>max</sub> and TBR to ROSEM (SUV<sub>max</sub>: <i>p</i> = 0.19–0.61; TBR: <i>p</i> = 0.26–0.70), and CNR values were comparable between DPR3 and ROSEM (<i>p</i> = 0.75). Regarding lesion volume, measurements from ROSEM reconstructions agreed closely with CT-based volumes (<i>p</i> = 0.48), whereas OSEM and all DPR groups consistently overestimated volumes (all <i>p</i> &lt; 0.05). DPR3 achieved the highest overall image quality score and was comparable to ROSEM (<i>p</i> = 0.85).</p> Conclusions <p>DPR algorithm significantly improves the quantification and visualization of small oncologic lesions compared to the conventional OSEM reconstruction. DPR with a filter strength factor of 3 (DPR3) offers the optimal balance, providing superior CNR and visual quality, although ROSEM remains superior for volumetric accuracy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Investigate the quantification accuracy of small lesions in oncological 18F-FDG PET/CT using a deep progressive learning reconstruction method

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月17日

查看全文
  • Lei Xu,
  • Rui Yang,
  • Ru-shuai Li,
  • Ren-cong Liu,
  • Qing-le Meng,
  • Feng Wang

摘要

Background

To investigate the impact of deep progressive learning reconstruction (DPR) on small lesion detection and image quality compared to ordered subset expectation maximum (OSEM) and regularized OSEM (ROSEM) in 18F-FDG PET/CT imaging.

Methods

The NEMA phantom was filled with 18F-FDG solution, with a hot sphere-to-background ratio of 4:1. Twenty-six patients with 18F-FDG-avid lung lesions (diameter < 2.0 cm) were enrolled in the study. The PET images were reconstructed by seven groups: routine OSEM, ROSEM with a penalization factor of 0.8 (ROSEM), and DPR reconstructions with five different filter strength factors ranging from smooth to sharp:1–5 (DPR1, DPR2, DPR3, DPR4, and DPR5). The contrast recovery (CR), background variability (BV), contrast-to-noise ratio (CNR), and radioactivity concentration ratio (RCR) were measured in the phantom study. The maximum standardized uptake value (SUVmax), target-to-background ratio (TBR), CNR, the volume of the lesions, and the coefficient of variation (COV) of the liver were calculated and compared between these methods in the patient study. Two radiologists evaluated the image quality using a five-point Likert scale.

Results

In the phantom study, the DPR2 to DPR5 and ROSEM groups achieved higher CR, CNR, RCR, and lower BV than the OSEM group. For the smallest 10-mm hot sphere, DPR3 achieved a CNR of 24.46, compared to 22.25 for ROSEM and 10.39 for OSEM. In patient studies, liver background noise (COV) was significantly lower in DPR1-DPR4 compared to OSEM (all p < 0.05), with no significant difference between the DPR3 and ROSEM (p = 0.65). Lesion quantification metrics-SUVmax, TBR, and CNR were significantly higher in DPR2-DPR4 than in OSEM (all p < 0.05). Notably, DPR4 and DPR5 provided equivalent lesion SUVmax and TBR to ROSEM (SUVmax: p = 0.19–0.61; TBR: p = 0.26–0.70), and CNR values were comparable between DPR3 and ROSEM (p = 0.75). Regarding lesion volume, measurements from ROSEM reconstructions agreed closely with CT-based volumes (p = 0.48), whereas OSEM and all DPR groups consistently overestimated volumes (all p < 0.05). DPR3 achieved the highest overall image quality score and was comparable to ROSEM (p = 0.85).

Conclusions

DPR algorithm significantly improves the quantification and visualization of small oncologic lesions compared to the conventional OSEM reconstruction. DPR with a filter strength factor of 3 (DPR3) offers the optimal balance, providing superior CNR and visual quality, although ROSEM remains superior for volumetric accuracy.