不同重建算法对肺功能CT定量参数的影响  

Exploring the effects of different reconstruction algorithms on CT quantitative parameters of lung function

作　　者：王旭[1] 刘义军[1] 赵明月 李贝贝 周宇婧 刘爱连[1] WANG Xu;LIU Yijun;ZHAO Mingyue;LI Beibei;ZHOU Yujing;LIU Ailian(Department of Radiology,the First Affiliated Hospital of Dalian Medical University,Dalian,Liaoning Province 116011,China)

机构地区：[1]大连医科大学附属第一医院放射科,辽宁大连116011

出　　处：《实用放射学杂志》2023年第8期1350-1353,共4页Journal of Practical Radiology

摘　　要：目的探讨滤波反投影(FBP)、Karl迭代和深度学习(DL)3种重建技术对肺功能CT定量参数的影响。方法选取行胸部CT体检有吸烟史者43例,对体检者胸部高分辨率CT(HRCT)扫描数据进行FBP、Karl迭代(1~9级)、DL(1~4级)算法图像重建,共获得14组图像,A组为FBP图像,B(B1~B9)组为Karl迭代图像,C(C1~C4)组为DL图像。所有重建图像传输至GE AW4.6工作站,利用ThoracicVCAR肺功能软件进行定量分析,获得肺内低密度占比(LAA%),全肺体积(TLV),同时手动选取右肺上叶最垂直的节段性支气管,生成从气管到节段性支气管的路径,自动显示气道管壁的平均厚度及管壁面积比(WA%)的测量结果。以FBP算法为对照,采用Friedman检验比较A、B、C组及B、C组组内LAA%、气道管壁的平均厚度及WA%的差异,采用单因素方差分析比较TLV的差异。结果A、B、C组及B、C组组内TVL、支气管管壁平均厚度、WA%差异无统计意义(P>0.05);A、B、C组间LAA%差异有统计学意义(χ2=525.562,P<0.001)。B组组内LAA%呈明显下降趋势;C组组内LAA%呈先略升高后稍下降趋势,其中C1组与A组间LAA%差异有统计学意义(P<0.05)。结论不同Karl迭代级别对于肺功能CT定量参数LAA%影响较大,临床应用时建议采用低权重;而不同DL级别数据一致性较好,均可用于肺功能定量分析。Objective To investigate the effects of three reconstruction techniques,filtered back projection(FBP),Karl iteration and deep learning(DL),on quantitative CT parameters of lung function.Methods A total of 43 cases underwent chest CT with a history of smoking were collected.The chest high resolution computed tomography(HRCT)scan data of the subjects were reconstructed with FBP,Karl iteration(levels 1-9),and DL(levels 1-4)algorithm to obtain 14 groups of images,with FBP images in group A,Karl iteration images in group B(B1-B9),and DL images in group C(C1-C4).All reconstructed images were transferred to the GE AW4.6 workstation for quantitative analysis using the Thoracic VCAR lung function software to obtain the low attenuation area ratio(LAA%),total lung volume(TLV),while the most vertical segmental bronchus in the upper lobe of the right lung was manually selected to generate a path from the trachea to the segmental bronchus and automatically display the average thickness of the airway wall and the wall area ratio(WA%).The FBP algorithm was used as a control,and the Friedman test was used to compare the differences in LAA%,mean thickness of the airway wall and WA%among the groups A,B and C,and between groups B and C,and a one-way ANOVA was used to compare TLV differences among the groups.Results The differences in TVL,mean bronchial wall thickness,and WA%were not statistically significant among the groups A,B,C,and between groups B and C(P>0.05).The differences in LAA%were statistically significant among the groups A,B,and C(x2=525.562,P<0.001).LAA%showed a significant decreasing in the group B,and a slight increasing followed by a slight decreasing in the group C,and there was statistical significance between the group Cl and the group A(P<0.05)..Conclusion Different Karl iteration levels have a greater impact on the lung function quantitative CT parameter of LAA%,and low weights are recommended for clinical applications.While different DL levels have better consistency,and can all be used for quantitative lung

关 键 词：肺功能 定量分析 滤波反投影 迭代重建 深度学习算法 

分 类 号：R322.35[医药卫生—人体解剖和组织胚胎学] R814.42[医药卫生—基础医学]