| 张磊,时洪坤,董书杉,等.人工智能图像优化技术在低剂量胸部CT检查中的初步应用研究[J].中华放射医学与防护杂志,2020,40(9):722-727.Zhang Lei,Shi Hongkun,Dong Shushan,et al.Impact of artificial intelligence imaging optimization technique on image quality of low-dose chest CT scan[J].Chin J Radiol Med Prot,2020,40(9):722-727 |
| 人工智能图像优化技术在低剂量胸部CT检查中的初步应用研究 |
| Impact of artificial intelligence imaging optimization technique on image quality of low-dose chest CT scan |
| 投稿时间：2020-06-03 |
| DOI：10.3760/cma.j.issn.0254-5098.2020.09.013 |
| 中文关键词: 人工智能 胸部计算机体层成像 辐射剂量 |
| 英文关键词:Artificial intelligence Chest computed tomography Radiation dose |
| 基金项目: |
|
| 摘要点击次数: 3023 |
| 全文下载次数: 1598 |
| 中文摘要: |
| 目的 探讨人工智能（AI）图像优化技术对低剂量胸部CT平扫图像质量及辐射剂量的影响。方法 前瞻性连续纳入2019年7月至8月于吉林大学第一医院采用NeuViz Prime CT行胸部CT平扫的80例患者，按随机数表法分为A、B两组，每组40例。A组为低剂量组，B组为常规剂量组，分别采用100及120 kV管电压；两组均采用自动管电流技术，参考毫安秒分别为70及140 mAs。根据重建方法的不同，将低剂量组分为A1、A2两个亚组，A1组为低剂量迭代组，采用迭代算法（ClearView 50%）重建图像；A2组为低剂量AI组，采用AI图像优化算法进一步优化A1组图像；B组采用迭代算法（ClearView 50%）重建图像。通过容积CT剂量指数（CTDIvol）、剂量长度乘积（DLP）和有效辐射剂量（E）的值，比较A、B两组辐射剂量的差异。比较A1、A2及B组感兴趣区的噪声值（SD）、信噪比（SNR）及对比噪声比（CNR）。由两名高年资放射科医生以Likert 5级评分法对3组图像质量进行主观评价。结果 A、B两组患者临床资料的比较差异均无统计学意义。A组与B组相比[（1.48±0.49）mSv vs.（5.30±1.40）mSv]，有效辐射剂量降低约72.1%。在图像质量方面，与B组相比，A1组SD较高且SNR及CNR较低（ZSD=-4.24，ZSNR=-2.54，tCNR=-2.27，P<0.05）。经AI优化后，A2组的SD显著低于B组（ZSD=-28.24，P<0.001），且SNR及CNR显著高于B组（tSNR=-26.04，tCNR=-36.88，P<0.001）；两组图像噪声的主观评分差异无统计学意义，但B组在肺内组织结构显示方面优于A2组（χ2=4.96、7.04，P<0.05）。结论 在辐射剂量降低约72.1%的情况下，经AI优化的低剂量胸部CT图像可达到常规剂量图像质量水平。 |
| 英文摘要: |
| Objective To investigate the impact of artificial intelligence imaging optimization technique on the image quality and radiation dose of low-dose chest CT scan. Methods Eighty patients who underwent chest CT examination in the Jilin University 1st hospital from July to August, 2019 were randomly divided into two groups(A, B), with 40 patients in each. The voltage of group A was 100 kV, while the other was 120 kV. According to different reconstruction method, group A was divided into two subgroups, group A1 and group A2. The images of A1 were reconstructed by iterative algorithm (ClearView 50%), while A2 images were optimized A1 by NeuAI imaging optimization technique. Group B used iterative algorithm (ClearView 50%) to reconstruct the image. The CT dose index (CTDIvol), dose-length product (DLP) and effective radiation dose (E) of group A and group B were recorded and compared.Objective the evaluation indicators were CT noise (SD), signal-to-noise ratio (SNR) and comparative noise ratio (CNR) of ROI. Subjective evaluation was done by 2 chief radiologists using double-blind method and image quality was graded by 5-point Likert scale.Results The patient characteristics between group A and group B showed no significant differences(P>0.05). Compared with group B, the effective radiation dose in group A was reduced by 72.1%[(1.48±0.49) mSv vs. (5.30±1.40) mSv]. The SD in group A1 was higher than that in group B, while SNR and CNR were lower (ZSD=-4.24, ZSNR=-2.54, tCNR=-2.27, P<0.05). The SD in group A2 was significantly lower than that in group B (ZSD=-28.24, P<0.001), and SNR and CNR were significantly higher than that in group B (tSNR=-26.04, tCNR=-36.88, P<0.001). There was no significant difference in subjective scores of image noise between group A2 and group B, while subjective scores of lung structure in group B were better than those in group A2(χ2=4.96、7.04,P<0.05). Conclusions Although the radiation dose was reduced by 72.1%, the low-dose chest CT images optimized by AI could reach the image quality level of standard dose. |
| HTML 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|