期刊文献+

深度学习重建算法在颞骨低剂量CT检查中的应用研究

Application of Deep Learning Reconstruction Algorithm in Low Dose Temporal Bone CT
下载PDF
导出
摘要 目的探究深度学习重建(Deep Learning Reconstruction,DLR)算法在低剂量(Low Dose,LD)颞骨CT检查中改善图像质量的应用价值。方法前瞻性收集2021年5—9月在北京协和医院行颞骨CT检查的患者98例。将这些患者采用随机分组法分为正常剂量(Routine Dose,RD)组和LD组,各49例。RD组采用滤波反投影(Filter Back Projection,FBP)算法重建,LD组分别用FBP和DLR算法进行图像重建,分别记为RD-FBP、LD-FBP和LD-DLR。客观图像质量分析采用测量4个目标结构的CT值及图像SD值的方法,并计算信号噪声比(Signal to Noise Ratio,SNR)。主观图像质量分析采用主观盲法对16个中耳和内耳解剖结构进行主观评分。结果RD-FBP、LD-FBP和LD-DLR 3种图像的CT值差异无统计学意义(均P>0.05)。RD-FBP的SD值明显低于LD-FBP,LD-DLR的SD值明显低于RD-FBP。RD-FBP的SNR明显高于LD-FBP,LDDLR的SNR明显高于RD-FBP。RD-FBP和LD-DLR的评分明显高于LD-FBP(均P<0.001),LD-DLR与RD-FBP评分差异无统计学意义(P>0.05)。结论DLR技术可以在降低颞骨CT辐射剂量的同时提高图像质量,满足临床诊断需求。 Objective To explore the application value of deep learning reconstruction(DLR)algorithm in low dose(LD)temporal bone CT to improve image quality.Methods Ninety-eight patients who underwent temporal bone CT in Peking Union Medical College Hospital from May to September 2021 were prospectively collected.These patients were randomly divided into routine dose(RD)group and low dose(LD)group,with 49 patients in each group.The RD group was reconstructed by filter back projection(FBP)algorithm,while the LD group was reconstructed by FBP and DLR algorithms respectively,and they were denoted as RD-FBP,LDFBP and LD-DLR respectively.Objective image quality analysis was performed by measuring the CT values and image SD values of the four target structures and calculating the signal to noise ratio(SNR).Subjective image quality analysis was performed by scoring sixteen middle and inner ear anatomical structures.Results CT values of RD-FBP,LD-FBP and LD-DLR images showed no statistical diference(all P>0.05).The SD value of RD-FBP was significantly lower than that of LD-FBP,and the SD value of LDDLR was significantly lower than that of RD-FBP.The SNR of RD-FBP was significantly higher than that of LD-FBP,and SNR of LD-DLR was significantly higher than that of RD-FBP.The image score of RD-FBP and LD-DLR were significantly higher than those of LD-FBP(both P<0.001),however,there was no significant diference between LD-DLR and RD-FBP(P>0.05).Conclusion DLR technology can reduce the radiation dose of temporal bone CT and improve the image quality to satisfy clinical diagnosis.
作者 王天娇 陈钰 王沄 王晓 王曼 许英浩 马壮飞 付海鸿 金征宇 WANG Tianjiao;CHEN Yu;WANG Yun;WANG Xiao;WANG Man;XU Yinghao;MA Zhuangfei;FU Haihong;JIN Zhengyu(Department of Radiology,Peking Union Medical College Hospital,Beijing 100730,China;Canon Medical System(China)Co.,Ltd.,Beijing 100015,China)
出处 《中国医疗设备》 2023年第1期93-97,121,共6页 China Medical Devices
关键词 深度学习重建 颞骨CT 低剂量 辐射剂量 图像质量 deep learning reconstruction temporal bone CT low dose radiation dose image quality
  • 相关文献

参考文献6

二级参考文献29

共引文献37

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部