DLIR与ASIR-V算法在70kVp低管电压行冠状动脉CT血管成像对图像质量的影响  

Comparison of image quality between DLIR and ASIR-V algorithm for coronary CT angiography using 70 kVp low tube voltage

作　　者：文雨婷 李万江[1] 潘雪琳[1] 帅桃[1] 彭礼清[1] 刘嘉丽 李真林[1] WEN Yuting;LI Wanjiang;PAN Xuelin;SHUAI Tao;PENG Liqing;LIU Jiali;LI Zhenlin(De partment of Radiology,West China Hospital of Sichuan University,Chengdu 610041,China)

机构地区：[1]四川大学华西医院放射科,四川成都610041

出　　处：《实用放射学杂志》2022年第6期994-997,1017,共5页Journal of Practical Radiology

基　　金：四川省科学技术厅重点研发项目(2019YFS0522)。

摘　　要：目的探讨深度学习图像重建(DLIR)与自适应统计迭代重建(ASIR-V)算法在70 kVp低管电压下行冠状动脉CT血管成像(CCTA)对图像质量的影响.方法前瞻性选取经临床医师评估拟行CCTA扫描且体质量指数(BMI)正常(18 kg/m^(2)≤BMI<24 kg/m^(2))的患者32例.选用管电压70 kVp,智能毫安调制,噪声指数(NI)为28;高浓度对比剂碘美普尔(400 mg I/mL),对比剂用量0.4 mL/kg.设备自动选择最佳时相后,利用CCTA的原始数据分别重建该最佳时相的高权重DLIR(DLIR-H)、中等权重DLIR(DLIR-M)、低权重DLIR(DLIR-L)以及50%ASIR-V的图像,共获得4组重建图像,比较4组图像质量之间的差异.采用5分法对图像质量进行主观评分(5=最佳);右冠状动脉(RCA)、左冠状动脉前降支(LAD)、左回旋支(LCX)的图像噪声、信噪比(SNR)及对比噪声比(CNR)用于客观图像质量评价.结果4组间DLIR-H组图像噪声值最低(12.68 HU±2.85 HU),50%ASIR-V组最高(21.98 HU±4.74 HU)(P<0.05);DLIR-H组RCA、LAD、LCX的SNR及CNR最高(35.78±12.23,34.51±11.48,34.79±11.91;47.43±13.94,46.16±13.07,46.44±13.56),50%ASIR-V组最低(20.03±6.41,19.47±6.21,19.41±6.14;26.80±7.21,26.25±7.04,26.19±6.95)(P<0.05);4组间图像CT值无明显差异(P>0.05).DLIR-H组图像主观评分最高,且与ASIR-V组图像主观评分差异有统计学意义(P<0.05).结论DLIR算法有助于在低千伏扫描条件下降低图像噪声,提高血管的SNR及CNR.在70 kVp低管电压行CCTA扫描时,ASIR-V算法与DLIR算法相比,DLIR-H算法图像质量主观评分更高,客观图像噪声更低,SNR和CNR更高.Objective To compare the image quality of deep learning image reconstuction(DLIR)and adaptive statistical iterative reconstruction(ASIR-V)algorithms in coronary computed tomography angiography(CCTA)using 70 kVp low tube voltage.Methods 32 normal body mass index(BMI)(18 kg/m^(2)≤BMI≤24 kg/m^(2))patients were prospectively enrolled to underwent CCTA.Tube voltage 70 kVp,intelligent milliampere modulation,noise index(NI)was 28.The highr concentration contrast agent(Iomeprol,400 mg I/mL)was used at a dosage of 0.4 mL,/kg.The original CCTA raw data was reconstructed at the best cardiac phase(recommended by the scanner)using DLIR-high(DLIR-H),DLIR-medium(DLIR-M),and DLIR-low(DLIR-L)strengths and 50% ASIR-V algorithms.A total of 4 image groups were obtained,and their differences in image quality were compared.The subjective image quality was assessed using a 5-point system(5=best)and compared with Friedman test.The image noise,signal-to noise ratio(SNR)and contrast to noise ratio(CNR)of right coronary artery(RCA),left anterior descending(LAD),left circumflex(LCX)were measured and compared with one way ANOVA.Results Among the 4 groups,DLIR-H had the lowest image noise value(12.68 HU±2.85 HU),while 50% ASIR-V had the highest(21.98 HU±4.74 HU)(P<0.05).DLIR-H also had the highest SNR(35.78±12.23,34.51±11.48,34.79±11.91)and CNR(47.43±13.94,46.16±13.07,46.44±13.56)of RCA,LAD and LCX,while the 50% ASIR-V had the lowest values(20.03±6.41,19.47±6.21,19.41±6.14 and 26.80±7.21,26.25±7.04,26.19±6.95,respectively)(P<0.05).There was no significant difference in image CT value among the 4 groups(P>0.05).The subjective scores in the DLIR-H ranked highest,followed by DLIR-M,DLIR-L and 50% ASIR-V and the differences between the DLIR groups and 50% ASIR-V were statistically significant(P<0.05).Conclusion DLIR algorithms significantly reduce image noise and improve SNR and CNR of vessels under low kV scanning conditions.Compared with the conventional ASIR-V algorithm,DLIR-H provides better subjective image score,lower imag

关 键 词：计算机体层成像 冠状动脉CT血管成像 图像质量 深度学习图像重建 

分 类 号：R814.42[医药卫生—影像医学与核医学] R543.3[医药卫生—放射医学] R814.49[医药卫生—临床医学]