多模型自适应统计迭代重建算法对鼻窦CT图像质量和辐射剂量影响的实验研究

Experimental study on the effect of adaptive statistical iterative reconstruction algorithms on image quality and radiation dose in paranasal sinus CT

目的探讨鼻窦CT扫描中,前置和后置自适应统计迭代重建算法(ASiR-V)对图像质量和辐射剂量的影响,并寻找最佳的迭代组合。方法以1具离体头颅标本为研究对象,采用临床鼻窦CT常规扫描条件[噪声指数(NI)=8],以及前置ASiR-V的不同等级(0~100%,间隔为10%)进行螺旋扫描,所得原始数据使用后置ASiR-V的不同等级(0~100%,间隔为10%)进行骨算法和标准算法重建,共获得242个鼻窦薄层图像序列。选择特定的感兴趣区(ROI)测量CT值,并计算图像对比噪声比(CNR)和品质因子(FOM)。记录容积CT剂量指数(CTDIvol)和智能毫安(Smart mA)值。采用线性回归分析对ASiR-V各等级与对应的CTDIvol、Smart mA、CNR、FOM进行比较分析。采用配对t检验对相同后置不同前置ASiR-V骨算法和标准算法图像的CNR进行分析比较。主观评价采用双盲法由3名高年资放射诊断医师以4分法(4分为最佳)进行图像质量评分。结果随着前置ASiR-V等级(0~100%)的增加,Smart mA、CTDIvol均减低,呈线性负相关(r分别为-0.981、-0.976,P均<0.001);Smart mA降幅为72.05%,CTDIvol降幅为71.22%。前置ASiR-V相同,随后置ASiR-V等级增加,骨算法和标准算法图像对应的CNR呈上升趋势,呈正相关(骨算法图像:R2分别为0.976、0.992、0.982、0.982、0.975、0.975、0.979、0.996、0.952、0.978、0.965;标准算法图像:R2分别为0.944、0.990、0.988、0.993、0.996、0.987、0.984、0.996、0.996、0.990、0.965);后置ASiR-V相同,随前置ASiR-V等级增加,骨算法和标准算法图像对应的FOM呈波动变化(骨算法图像:R2分别为0.335、0.341、0.344、0.364、0.385、0.405、0.418、0.429、0.455、0.474、0.516;标准算法图像:R2分别为0.223、0.278、0.327、0.285、0.309、0.329、0.325、0.346、0.360、0.390、0.380)。以上各种前置和后置迭代等级组合所得图像主观评价均可满足诊断要求(评分≥3)。结论当NI=8时,骨算法最佳前置和后置迭代等级组合为80%和100%;标准算法最佳迭代等级组合为100%和100%。鼻窦CT扫描中,选择恰当的前置和后置迭代等级组合,能够在图像质量满足诊断要求的前提下,有效降低辐射剂量。

Objective To explore the effect of pre-and post-adaptive statistical iterative reconstruction-Veo(ASiR-V)on image quality and radiation dose in paranasal sinus CT,and to find the best combinations.Methods One head specimen was scanned with the routine spiral CT scanning parameters[noise index(NI)=8]and different levels of pre-ASiR-V(0—100%,with an interval of 10%).The images were reconstructed with different post-ASiR-V(0—100%,with an interval of 10%)for the bone algorithm and standard algorithm.All of 242 thin-layer images of paranasal sinuses were obtained.The region of interest(ROI)was selected to measure the CT value to calculate the contrast to noise ratio(CNR)and figure of merit(FOM).The volume CT dose index(CTDIvol)and Smart mA were recorded.The linear regression was conducted to analyze the relationship between CTDIvol,SmartmA,CNR and FOM.And with the same pose-ASiR-V level,the CNR of images which reconstructed by bone and soft algorithms were compared with pair-wise t test.The image quality was subjectively evaluated by three independent experienced radiologists using a 4-point scale(4 was the best).Results As the pre-ASiR-V levels(0—100%)increased,Smart mA and CTDIvol were reduced with a linear negative correlation(r=-0.981,-0.976,both P<0.001).The Smart mA decreased by 72.05%and CTDIvol by 71.22%.Keeping the same pre-ASiR-V level,the CNR increased with the increase of post-ASiR-V level(for the bone algorithm images:R2=0.976,0.992,0.982,0.982,0.975,0.975,0.979,0.996,0.952,0.978,0.965;for the standard algorithm images:R2=0.944,0.990,0.988,0.993,0.996,0.987,0.984,0.996,0.996,0.990,0.965).Under the same level of post-ASiR-V,the CNR and FOM fluctuated with the pre-ASiR-V level(for the bone algorithm images:R2=0.335,0.341,0.344,0.364,0.385,0.405,0.418,0.429,0.455,0.474,0.516;for the standard algorithm images:R2=0.223,0.278,0.327,0.285,0.309,0.329,0.325,0.346,0.360,0.390,0.380).All subjective image quality could meet the diagnostic requirements(the score≥3).Conclusion At NI=8,for the bone algorithm,the best combination is 80%pre-ASiR-V and 100%post-ASiR-V;for the standard algorithm,the best iteration combination is 100%and 100%.The appropriate choice of pre-and post-ASiR-V levels in paranasal sinus CT scan can effectively reduce the radiation dose under the premise of maintaining the image quality that meets the diagnostic needs.