迭代重建技术在小猪胸部容积CT扫描中的应用及最佳剂量研究

The optimal dose reduction level in chest CT with CT volume scan mode using iterative reconstruction in piglet model

目的 探讨迭代重建技术（AIDR 3D）在小猪胸部CT容积扫描中的可行性及维持图像质量的最适低剂量水平.方法 按小儿成长体质量标准表,依据0～3岁小儿体质量选择29只普通小猪进行640层CT胸部容积扫描.管电压80 kVp,采用sureExposure3D技术,每只小猪均进行常规剂量[噪声指数设置为10.0,A组]及4级不同水平的低剂量（噪声指数分别为12.5、15.0、17.5、20.0,分别记作B、C、D、E组）扫描.A组采用滤波反投影重建,各低剂量组均采用AIDR 3D strong重建.测量每组图像的客观噪声及信噪比,并由2名放射诊断医师采用盲法对肺窗图像（伪影、中心气道显示及肺组织显示）及纵隔窗图像（纵隔结构显示）及整体主观图像质量进行5分法评分.用随机区组设计方差分析比较各组的CT值、客观噪声及信噪比及辐射剂量（CT容积剂量指数、CT剂量长度乘积、有效辐射剂量）的差异,并采用LSD检验进行两两比较;用Friedman检验比较各组主观图像质量评分的差异,并采用Wilcoxon检验进行两两比较.采用ROC曲线分析获得可维持良好图像质量（评分≥4分）的最适低剂量水平.结果 与A组[噪声（15.25 ±3.14）HU、信噪比3.23±1.07]比较,B、C组噪声[（12.11±2.75）、（13.18 ±3.16）HU]明显降低（F=7.38,P＜0.01）,信噪比（4.13±1.38、3.80±1.20）明显提高（F =3.11,P＜0.05）,差异均有统计学意义.D、E组的噪声[（15.14±4.51）、（15.79 ±4.17） HU]及信噪比（3.40±1.56、3.45±1.70）与A组比较差异无统计学意义（P值均＞0.05）.B、C组图像质量各主观指标评分优于A组（P值均＜0.01）,D组的主观图像质量评分与A组差异无统计学意义（p＞0.05）,而E组有5只小猪的整体图像质量评分＜3分.总体图像质量的ROC曲线表明D组是可以维持良好图像质量的最适剂量水平（曲线下面积0.75,95％可信区间0.58 ～0.92）.D组[（0.47 ±0.16） mGy]在维持与A组[（1.14 ±0.27） mGy]相当的主、客观图像质量的同时可使辐射剂量降低59％ （F=183.83,P＜0.01）.结论 在小猪行640层CT胸部容积扫描,采用AIDR 3D迭代重建结合sureExposure3D技术,在噪声指数设置为17.5的条件下,可获得与常规剂量组相当的图像质量,同时使辐射剂量降低59％.

Objective To determine the feasibility and the optimal dose reduction of AIDR 3D on piglet chest CT which can provide image quality （IQ） comparable to filtered back projection （FBP）.Methods Twenty-nine normal pigs with the weight of 3-12kg underwent 640-slice MDCT chest CT （Aquilion one,Toshiba） for 5 times with 80 kvp and various mAs.SureExposure3D technique were used and the index of noise were set to SD 10.0,12.5,15.0,17.5,20.0（Group A,B,C,D,E） to reduce dose successively.Group A were reconstructed with FBP,Group B,C,D,E were reconstructed using AIDR 3D （strong level）.Quantitative image noise and signal to noise ratio（SNR） were measured in each group.Two radiologists graded subject image quality on both lung images （artifacts,central airway,lung tissue） and mediastinal images （mediastinal structure） and overall image quality using a 5-point scale in a blinded manner.Analysis of variance （ANOVA） and LSD test were used for comparisons of objective evaluation indices （CT value,noise,SNR）and radiation dose（CTDIvo1,DLP,ED） among the five groups.The Friedman test and Wilcoxon test were used for comparisons of demographic data and for detection of differences in subjective evaluation of IQ among groups.A receiver-operating characteristic （ROC） analysis was performed to establish a radiation reduction threshold up to which comparable IQ（score≥4） was maintained.Results Compared with Group A[noise（15.25 ± 3.14）HU,SNR 3.23 ± 1.07],Group B,C had significant lower noise and higher SNR [noise （12.11 ± 2.75）,（13.18 ± 3.16）HU; SNR 4.13 ± 1.38,3.80 ± 1.20;F =7.38,3.11,P ＜0.05].Group D,E showed no significantly different noise and SNR with Group A[noise （15.14±4.51）,（15.79 ±4.17）HU;SNR 3.40± 1.56,3.45 ± 1.70;P＞0.05].Group B,C had significant better subject image quality compared with Group A （P ＜0.01）.Group D had no significantly different subject image quality compared with Group A （P ＞ 0.05）.Group E had 5 pigs with overall image scores of ＜ 3.The ROC curve of IQ established SD 17.5 （Group D） as optimal cut-off point （AUC 0.75,95％ CI 0.58-0.92）.Group D provided equivalent subjective image score and objective IQ measurements compared with FBP images in Group A.The ED of group D was 59％ lower than that of group A[CTDIvol （1.14±0.27） mGy vs （0.47 ±0.16）mGy,F=183.83,P＜0.01].Conclusion Using AIDR 3D technique,80 kvp with SureExposure3D （SD 17.5） can provide comparable IQ compared with routine dose with FBP reconstruction,and reduce 59％ dose in piglet model.