自适应统计迭代重建-V算法和深度学习图像重建算法对慢性阻塞性肺疾病（COPD）患者在超低剂量条件下的图像质量及肺气肿定量分析的影响

Effect of adaptive statistical iterative reconstruction-V algorithm and deep learning image reconstruction algorithm on image quality and emphysema quantification in COPD patients under ultra-low-dose conditions.

作者信息

Ma Guangming, Dou Yuequn, Dang Shan, Yu Nan, Guo Yanbing, Han Dong, Jin Chenwang

机构信息

Department of Radiology, The First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, Shannxi 710061, China.

Department of Radiology, Affiliated Hospital of Shaanxi University of Chinese Medicine, Xianyang, Shannxi 712000, China.

出版信息

Br J Radiol. 2025 Apr 1;98(1168):535-543. doi: 10.1093/bjr/tqae251.

DOI:10.1093/bjr/tqae251

PMID:39862404

Abstract

PURPOSE

To explore the effect of different reconstruction algorithms (ASIR-V and DLIR) on image quality and emphysema quantification in chronic obstructive pulmonary disease (COPD) patients under ultra-low-dose scanning conditions.

MATERIALS AND METHODS

This prospective study with patient consent included 62 COPD patients. Patients were examined by pulmonary function test (PFT), standard-dose CT (SDCT) and ultra-low-dose CT (ULDCT). SDCT images were reconstructed with filtered-back-projection (FBP), while ULDCT images were reconstructed using FBP, 30%ASIR-V, 60%ASIR-V, 90%ASIR-V, low-strength (DLIR-L), medium-strength (DLIR-M) and high-strength DLIR (DLIR-H) to form 8 image sets. Images were analysed using a commercial computer aided diagnosis (CAD) software. Parameters such as image noise, lung volume (LV), emphysema index (EI), mean lung density (MLD) and 15th percentile of lung density (PD15) were measured. Two radiologists evaluated tracheal and pulmonary artery image quality using a 5-point scale. Measurements were compared and the correlation between EI and PFT indices was analysed.

RESULT

ULDCT used 0.46 ± 0.22 mSv in radiation dose, 93.8% lower than SDCT (P < .001). There was no difference in LV and MLD among image groups (P > .05). ULDCT-ASIR-V90% and ULDCT-DLIR-M had similar image noise and EI and PD15 values to SDCT-FBP, and ULDCT-DLIR-M and ULDCT-DLIR-H had similar subjective scores to SDCT-FBP (all P > .05). ULDCT-DLIR-M provided the best correlation between EI and the FEV1/FVC and FEV1% indices in PFT, and the lowest deviations with SDCT-FBP in both EI and PD15.

CONCLUSION

DLIR-M provides the best image quality and emphysema quantification for COPD patients in ULDCT.

ADVANCES IN KNOWLEDGE

Ultra-low-dose CT scanning combined with DLIR-M reconstruction is comparable to standard dose images for quantitative analysis of emphysema and image quality.

摘要

目的

探讨在超低剂量扫描条件下，不同重建算法（自适应统计迭代重建-V（ASIR-V）和深度学习迭代重建（DLIR））对慢性阻塞性肺疾病（COPD）患者图像质量和肺气肿定量分析的影响。

材料与方法

本前瞻性研究经患者同意，纳入62例COPD患者。患者接受肺功能测试（PFT）、标准剂量CT（SDCT）和超低剂量CT（ULDCT）检查。SDCT图像采用滤波反投影（FBP）重建，而ULDCT图像采用FBP、30%ASIR-V、60%ASIR-V、90%ASIR-V、低强度（DLIR-L）、中等强度（DLIR-M）和高强度DLIR（DLIR-H）重建，形成8组图像。使用商用计算机辅助诊断（CAD）软件对图像进行分析。测量图像噪声、肺容积（LV）、肺气肿指数（EI）、平均肺密度（MLD）和肺密度第15百分位数（PD15）等参数。两名放射科医生使用5分制评估气管和肺动脉图像质量。对测量结果进行比较，并分析EI与PFT指标之间的相关性。

结果

ULDCT的辐射剂量为0.46±0.22 mSv，比SDCT低93.8%（P<.001）。各图像组之间LV和MLD无差异（P>.05）。ULDCT-ASIR-V90%和ULDCT-DLIR-M的图像噪声、EI和PD15值与SDCT-FBP相似；ULDCT-DLIR-M和ULDCT-DLIR-H的主观评分与SDCT-FBP相似（均P>.05）。ULDCT-DLIR-M在PFT中EI与第1秒用力呼气容积/用力肺活量（FEV1/FVC）和第1秒用力呼气容积占预计值百分比（FEV1%）指标之间的相关性最佳，且EI和PD15与SDCT-FBP的偏差最小。