摘要
目的探讨深度学习图像重建(DLIR)算法、多模型迭代重建(ASiR-V)算法和滤波反投影(FBP)算法在低管电压条件下对腹部平扫CT图像质量的影响。方法前瞻性搜集因病情需要行全腹部CT平扫检查的56例患者,根据体质量指数(BMI)将入选患者分为A组(18 kg/m^(2)≤BMI<24 kg/m^(2),管电压80 kVp,n=29)、B组(24 kg/m^(2)≤BMI<29 kg/m^(2),管电压100 kVp,n=27)。所有图像数据均进行FBP、权重为50%的ASiR-V(ASiR-V50%)和高强度DLIR(DLIR-H)图像重建。采用Kruskal-Wallis H检验比较不同重建算法图像间的各项客观评价指标[噪声、肝脏和胰腺的信号噪声比(SNR)、肝脏和胰腺的对比噪声比(CNR)]和主观评价指标(噪声、总体图像质量),组内两两比较采用Bonferroni校正检验。结果A组和B组的DLIR-H图像的噪声,肝脏、胰腺SNR均显著优于FBP和ASiR-V50%,差异均有统计学意义(P值均<0.05)。在80 kVp组和100 kVp组,DLIR-H的噪声较FBP降低66.8%和68.7%,较ASiR-V50%降低46.1%和48.7%。DLIR-H肝脏和胰腺的CNR高于FBP,差异有统计学意义(P<0.05),DLIR-H和ASiR-V50%、ASiR-V50%和FBP间肝脏和胰腺的CNR差异无统计学意义(P值均>0.05)。主观评分上,A组和B组DLIR-H的主观噪声和总体图像质量高达4分以上,均高于FBP和ASiR-V50%(P值均<0.05)。结论与ASiR-V50%和FBP相比,DLIR-H降低了图像噪声,提高了图像质量,在辐射剂量优化方面有更大的潜力。
Objective To investigate the effects of deep learning image reconstruction(DLIR)algorithm,adaptive statistical iterative reconstruction-V(ASiR-V)algorithm and filtered back projection(FBP)on abdominal plain CT image quality under low tube voltage.Methods Fifty-six patients who underwent abdominal CT plain scans were collected prospectively and divided into group A(18 kg/m^(2)≤BMI<24 kg/m^(2),80 kVp,n=29)and Group B(24 kg/m^(2)≤BMI<29 kg/m^(2),100kVp,n=27)according to their BMI.All images were reconstructed with FBP,ASiR-V50%and DLIR at highstrength level(DLIR-H).Quantitative parameters[image noise,liver and pancreas signal to noise ratio(SNR),liver and pancreas contrast to noise ratio(CNR)] and qualitative parameters(image noise and overall image quality)were compared by Kruskal-Wallis H test,and the Bonferroni test was used for multiple comparisons within the group.Results The noise,liver,and pancreas SNR of DLIR-H images were significantly better than those of FBP and ASiR-V50%in both group A and group B.The differences were statistically significant(all P<0.05).In the 80 kVp and 100 kVp groups,the noise of DLIR was reduced by 66.8%and 68.7%compared with FBP,and by 46.1%and 48.7%compared with ASiR-V50%.The CNR of liver and pancreas of DLIR-H was higher than that of FBP,and the dfferences were statistically significant(P<0.05).There was no significant difference in CNR of liver and pancreas between DLIR-H and ASiR-V50%,ASiR-V50%and FBP(both P>0.05).On subjective scores,DLIR-H in groups A and B had subjective noise and overall image quality up to 4 points,both higher than FBP and ASiR-V50%.Conclusion Compared with ASiR-V50%and FBP,DLIRH reduced noise and improved overall image quality which may promote a further radiation dose reduction in the future.
作者
刘娜娜
吕培杰
王会霞
刘星
詹鹏超
陈岩
李臻
高剑波
LIU Nana;LYU Peijie;WANG Huixia(Department of Radiology,the First Affiliated Hospital of Zhengzhou University,Zhengzhou,Henan Province 450052,P.R.China)
出处
《临床放射学杂志》
北大核心
2023年第11期1820-1825,共6页
Journal of Clinical Radiology
基金
河南省高等学校重点科研项目(编号:22A320057)。
关键词
深度学习图像重建
低管电压
迭代重建算法
辐射剂量
图像质量
eep learning image reconstruction
Low tube voltage
Adaptive statistical iterative reconstruction-V
Radiation dose
Imagequality
