自适应迭代降剂量技术在个性化超低剂量胸部CT成像的应用性研究

Application of adaptive iterative dose reduction technique in personalized ultra low dose chest CT imaging

目的迭代算法是实现CT低剂量成像的有效工具,其特点是在降低图像噪声的情况下仍然保持较好的图像质量。但对于扫描条件过低噪声过大的图像,过度的迭代重建有可能降低图像的空间分辨率并损失解剖结构的细节。因此,如何能针对不同病人实施个性化成像、使放射剂量达到可处理的最低水平是在临床实施CT低剂量成像需要解决的问题。本文旨在探讨应用自适应迭代降剂量技术(adaptive iterative dose reduction 3D,AIDR3D)进行个体化超低剂量胸部CT成像的可行性。方法从本院2011年12月～2012年2月进行过首次常规剂量胸部CT平扫检查并且明确诊断的病人中,连续收集需要在1个月内进行CT平扫复查的病人共48例,其中男25例,女23例,13～84岁,平均年龄48.27±17.63岁,BMI为15.62～30.85,平均21.62±3.38。这48例病人首次常规剂量检查时采用普通自动曝光控制技术(automatic exposure control,AEC)扫描,其目标噪声值SD为12.5,并用常规滤波反投影(filter back projection,FBP)算法进行重建。复查时进行个性化超低剂量胸部CT成像方案扫描,扫描条件采用整合迭代算法(AIDR 3D)的AEC技术,设置目标噪声值为SD25,其他扫描参数和重建参数保持和常规检查一致,得到的数据分别进行AIDR算法和FBP算法重建。两次检查总共得到三组数据,分别为常规剂量FBP组(A组)、低剂量AIDR组(B组)和低剂量FBP组(C组)。3组分别用相应的过滤函数显示肺窗和纵隔窗。对三组数据进行图像质量评价并进行对比:客观评价用图像噪声值进行量化(用CT值的标准差表示),主观评价由两名医师采用3分制(3-优,2-可,1-差)进行独立盲法评分。为了评估结合了AIDR 3D的AEC技术调控管电流的效果,把胸部分成上、中、下部分别进行评价。数据统计采用随机区组设计的Friedman检验,将上、中、下评分均≥2分的病例定义为可诊断,计算可诊断率。扫描的放射剂量则通过记录机器上显示的CTDI和DLP数值,计算有效剂量ED(k=0.014),将复查和首次扫描的剂量进行对比。结果客观图像质量评价3组肺窗图像的噪声值均有统计学差异(P<0.05),B组比C组大幅降低噪声值(上66.58%;中39.62%;下48.55%),而B组和A组相比,上肺和下肺噪声分别降低15.67%和15.26%,而中肺噪声值则增加9.33%。3组的纵隔窗图像的噪声值相比均有统计学差异(P<0.05),B组比C组噪声值显著降低(上66.58%:中39.62%;下48.56%),而B组比A组噪声值有轻度增加(上39.95%;中79.43%;下76.35%)。主观图像质量评价:B组的肺窗上、中、下部之间评价无统计学差异(P>0.05),纵隔窗的上、中、下评价之间无统计学差异(P>0.05)。3组的肺窗图像的可诊断率(上、中、下均在2分或以上)都达到100%,优、良、差的分布没有统计学差异(P>0.05),但3组纵隔窗图像的可诊断率有显著差异,C组的可诊断率明显较低,B组和A组的可诊断率相同(95.83%vs 95.83%vs 56.25%,P<0.05)。放射剂量方面,复查扫描(B/C组)和初次扫描(A组)相比有效剂量降低87.05%(0.715 vs 5.524 mSv,k=0.014)。结论采用整合AIDR 3D的AEC技术可以实现对不同体型的人群的胸部个性化超低剂量CT成像。

Objective To investigate the feasibility of Adaptive Iterative Dose Reduction （AIDR 3D） technique in ultra low dose CT chest imaging. Methods 48 patients （ 25 males and 23 females, aged 13-84 years old, BMI 15.62- 30.85 ） who needed follow-up CT exam were collected in the study. Initial scan of these 48 patients applied conventional automatic exposure control （AEC） technique with the target image noise of SD 12.5 and filter back projection （FBP） algorithm reconstruction. The follow-up exam applied personalized ultra low dose chest CT scan protocol with an innovative AEC technique （sureExposure） integrated with iterative algorithm （AIDR 3D）, with the target noise of SD 25 and both AIDRand FBP algorithm reconstruction. According to the filter kernels, the data were divided into conventional dose scan with FBP reconstruction （Group A）, low dose scan with AIDR reconstruction （Group B） and low dose scan with FBP reconstruction （Group C）. Image quality of the three groups was evaluated and compared including objective evaluation using image noise and subjective evaluation which was executed by two radiologists with 3-grade scale （3-good,2- acceptable, 1-bad）. The image was defined as diagnosable when the score of all parts〉2. The effect of AEC technique integrated with AIDR 3D on adapting tube current were evaluated in different parts of the chest （upper chest, middle chest and lower chest） with Friedman test. Effective dose was calculated with the CTDI and DLP values. Results Significant differences were found in objective evaluation of lung window images between each two groups （P 〈 0.05 ）, image noise of Group B was substantially lower than Group C （upper chest 66.58%, middle chest 39.62%, lower chest 48.55%）, though compared with Group A, image noise was reduced 15.67% and 15.26% in the upper and lower chest, respectively） while increasing 9.33% in middle chest. Of the mediastinum window images, significant differences were also found （P 〈 0.05 ）, image noise of Group B decreased when comparing with Group C （upper 66.58%, middle 39.62%, lower 48.56%） while increased when comparing with Group A （upper 39.95%, middle 79.43%, lower 76.35% ）. In subjective evaluation, no significant differences were found between upper, middle and lower chest in both lung window and mediastinum window images （P 〉 0.05）. Lung window images of all the groups achieved 100% diagnosable rate without significant differences in the distribution of good, acceptable and had images （P 〉 0.05）. There were significantly different in diagnosable rate of mediasinum images among the three groups. The diagnosable rate of Group C was much lower than that of Group B and Group A （56.25% vs 95.83% vs 95.83% ,P 〈 0.05）. The radiation dose of follow-up exam was 87.05% lower than the initial exam （0.715 vs 5.524 roSy, k=O.014）. Conclusion Personalized ultra low dose chest CT applying AEC technique integrated with AIDR 3D is feasible.