| 徐亚运,胡政杨,吕品,等.双层探测器光谱CT虚拟平扫应用于肺结节人工智能诊断的可行性研究[J].中华放射医学与防护杂志,2023,43(10):827-832.Xu Yayun,Hu Zhengyang,Lyu Pin,et al.Feasibility of artificial intelligence diagnosis of pulmonary nodules on virtual non-contrast images derived from dual-layer spectral detector CT[J].Chin J Radiol Med Prot,2023,43(10):827-832 |
| 双层探测器光谱CT虚拟平扫应用于肺结节人工智能诊断的可行性研究 |
| Feasibility of artificial intelligence diagnosis of pulmonary nodules on virtual non-contrast images derived from dual-layer spectral detector CT |
| 投稿时间：2023-02-24 |
| DOI：10.3760/cma.j.cn112271-20230224-00052 |
| 中文关键词: 计算机体层成像 人工智能 肺结节 辐射剂量 虚拟平扫 |
| 英文关键词:Computed tomography Artificial intelligence Lung nodule Radiation dosage Virtual non-contrast |
| 基金项目: |
|
| 摘要点击次数: 2024 |
| 全文下载次数: 905 |
| 中文摘要: |
| 目的 探讨利用双层探测器光谱CT肺部虚拟平扫(VNC)图像进行肺结节人工智能(AI)诊断的可行性。方法 回顾性分析2022年5-11月间,在南京大学医学院附属鼓楼医院行胸部CT平扫联合双期增强扫描的52例患者资料。选择静脉期扫描数据重建肺窗VNC图像。用AI诊断软件对常规平扫(TNC)和VNC图像进行分析,测量肺实质的CT值和噪声并计算信噪比(SNR),并由2名放射医师对图像质量进行主观评分。记录各期扫描的剂量长度乘积(DLP),并计算有效剂量(E)。采用配对t检验比较2组图像质量客观指标和辐射剂量,采用Wilcoxon非参数检验比较图像质量主观评分。采用Wilcoxon符号秩检验比较2组图像AI诊断的敏感性和假阳性检出率(FPDR)。结果 与TNC相比,静脉期VNC图像的噪声降低了13.8%,SNR升高14.9%,DLP和E均降低了33.3%,差异具有统计学意义(t=5.82、-5.35、22.93、22.92,P<0.05)。2组图像肺实质的CT值及主观评分差异均无统计学意义(P >0.05)。对于不同类型的肺结节,2组图像AI 诊断的敏感性差异均无统计学意义(P >0.05)。但对于直径≤4 mm 实性结节和全部肺结节总体而言,VNC组的FPDR略有升高,差异具有统计学意义(Z=-2.03、 -3.09,P<0.05),对于其他类型的肺结节,FPDR差异则无统计学意义(P >0.05)。结论 基于光谱CT的肺部静脉期VNC图像,在保证图像质量和肺结节AI诊断准确性,且FPDR没有显著升高的情况下,大幅降低患者辐射剂量,可以替代TNC进行常规应用。 |
| 英文摘要: |
| Objective To evaluate the feasibility of artificial intelligence (AI) diagnosis of pulmonary nodules on virtual non-contrast(VNC) images derived from dual-layer detector spectral CT. Methods Totally 52 patients who underwent non-contrast and dual-phase enhanced chest CT scan from May 2022 to November 2022 were enrolled in this study. The VNC images of lung were reconstructed based on venous phase data. CT values and image noise of lung parenchyma, signal-to-noise ratio (SNR) were measured. The dose-length product (DLP) of each scan was recorded and the effective dose (E) was calculated. All of the objective indicators of image quality and radiation dose were compared by Paired t test. Image quality was evaluated subjectively by two radiologists and compared with Wilcoxon non-parametric test. Wilcoxon symbolic rank test was used to compare the sensitivity and false positive detection rate (FPDR) of AI diagnosis between two groups. Results Compared with TNC, the noise of venous VNC image was decreased by 13.8%, SNR increased by 14.9%, and both of DLP and E decreased by 33.3% (t=5.82, -5.35, 22.93, 22.92,P <0.05). There were no significant differences in CT values and subjective scores between 2 groups (P >0.05). For different types of pulmonary nodules, there was no statistical difference in the sensitivity of AI diagnosis between two groups (P >0.05). For solid nodules with diameter ≤4 mm and all pulmonary nodules in general, FPDR in VNC group was slightly increased with statistical significance (Z=-2.03, -3.09,P<0.05), while for other types of pulmonary nodules, there was no statistical difference (P >0.05). Conclusions The VNC images of thoracic venous phase based on spectral CT can significantly reduce the radiation dose of patients while the image quality and the AI diagnostic sensitivity of pulmonary nodules remain unchanged, and the FPDR without significantly increase. And it could replace TNC for daily routine. |
| HTML 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|