【作者】 李琼；

【导师】 刘士远； 于红； 范丽；

【作者基本信息】 第二军医大学 ， 影像医学与核医学（专业学位）， 2013， 博士

【摘要】 【研究背景】在胸部疾病的多种检查手段中，CT无疑是最敏感、最有价值的检查方法之一。CT具有很高的密度分辨率，且无组织重叠等优点，较胸部X线片而言，大大提高了病变检出的敏感性和特异性。随着多层螺旋CT的发展及广泛普及，其应用数量急剧增加，据资料统计，2007年美国CT检查次数较1980年增长了23倍，但常规胸部CT的辐射剂量大约是胸片的100倍。美国有高达2%的癌症病例可能是由于CT扫描的辐射造成的[1]。CT电离辐射的致癌风险以及放射防护与安全问题开始成为关注的焦点。因此当前世界CT研究的重点是如何在获得满足诊断要求图像的前提下，尽可能使受检者接受的辐射剂量减低。美国国家癌症研究所（NLST）研究发现，低剂量CT筛查相比于胸片其检出率高，能够检出更多的肺结节和肺癌（包括早期肺癌），研究结果表明，低剂量CT筛查组和胸部X片筛查组中阳性率分别为24.2%和6.9%。可以降低被检者肺癌相关病死率20%[2]。然而，过度的降低剂量，又会导致相应的图像噪声的升高和对病灶诊断信心的降低。既往研究[3,4]表明，合理的CT参数对fGGO的客观显示非常重要，在一定程度上低剂量CT并不能完全反映fGGO的特征，辐射剂量过低，图像噪声增大，容易造成假象或漏诊，尤其对部分有基础肺疾病的人群，如：陈旧性肺结核、肺气肿、肺纤维化等。 Li[3]等研究83例病理学证实的肺癌中，39％的fGGO因为能见度低被误诊。Funama等[4]通过仿真胸部磨玻璃密度结节体模研究发现，管电流降至21、45mAs时，CT值为-800HU或更低密度的模拟GGO结节漏检率约为60%、36%，CT值高于-650HU的模拟GGO结节漏检率分别为17%、8%，因此低剂量CT应用于肺部磨玻璃密度影的检出有一定限度。传统滤波反投影（filtered back projection, FBP)是目前大多数商用CT机型所采取的标准CT重建技术。但它对噪声和伪影较为敏感，因此限制了辐射剂量的降低。为弥补噪声升高而导致的图像质量下降，各厂商分别新开发出各种图像后处理技术，比如GE公司的ASIR、Siemens的IRIS、Philips的iDose等，其核心均是迭代重建技术（Iterative reconstruction, IR）。国内外也有研究者采用IR技术对低剂量扫描的原始数据进行重建，研究发现，迭代重建技术有明显降低辐射剂量，并使图像噪声减少的特点[5-10]。本研究通过仿真胸部体模及临床对照研究相结合的方法探讨迭代重建技术在低剂量胸部CT中应用的潜在价值，摸索一套适合肺内GGO筛查的低剂量方案。第一部分迭代重建技术对低剂量胸部CT图像质量影响的仿真体模研究1.1辐射剂量和迭代重建技术对肺部HRCT图像CT值影响的体模研究【研究目的】评价辐射剂量和迭代重建技术iDose对肺部HRCT图像CT值的影响。【材料与方法】应用Philips Brilliance128排螺旋CT在5种不同的剂量水平（管电压120kV不变，管电流200mAs、100mAs、50mAs、25mAs、10mAs）下对仿真胸部体模进行HRCT扫描，并分别采用传统的滤波反投影（FBP）和迭代重建算法iDose（水平分别选4和7）进行重建。根据CT机器测定的CT剂量容积指数（CTDIvol）、剂量长度乘积（DLP）计算每次扫描的有效辐射剂量（ED），测量、记录不同剂量水平、不同算法下的模型肺实质的CT值和SD值并进行统计学分析。【结果】不同管电流、不同重建算法条件下测得的肺实质CT值之间无明显差异（p值均>0.05）。SD值在不同管电流条件下有明显的统计学意义（p值均<0.05），且管电流越小，SD值越大。在同一剂量水平下，SD值在不同重建算法间有明显的统计学意义（p值均<0.05），迭代重建算法iDose可以明显降低SD值，且iDose所选的水平越高，SD值越小。【结论】低剂量HRCT扫描可以准确的测量肺实质的CT值，同时随着剂量的降低，噪声明显增大。和传统的滤波反投影（FBP)相比，迭代重建算法iDose可以明显的改善噪声,且不同的iDOSE水平降低噪声的能力不一。1.2应用迭代重建技术的低剂量HRCT评估肺磨玻璃密度结节可行性的体模研究【研究目的】评价应用不同水平iDose迭代重建技术(Iterative reconstruction, IR)在低剂量HRCT中评估肺纯磨玻璃密度结节（Pure ground glass nodule, pGGN）的可行性。【材料与方法】应用Philips Brilliance128排螺旋CT在5种不同的剂量水平（管电压120kV不变，管电流200mAs、100mAs、50mAs、25mAs、10mAs）下对含有9个5~12mm的pGGN的仿真胸部体模进行HRCT扫描，并分别采用传统的滤波反投影（FBP）和iDose迭代重建技术（水平分别选4和7）进行数据重建。由两名有经验的影像科医师采用盲法对不同剂量水平、不同重建技术下显示的pGGN的图像质量进行主观评价，并用Kappa检验评价观察者间的一致性，测量并比较每次扫描的有效辐射剂量（ED）。【结果】对不同剂量水平、不同重建技术下显示的pGGN的图像质量进行主观评价，两名观察者间的一致性中等或较好（k>0.4）。在管电流低至10mAs水平时，各组pGGN均能显示，结合iDose迭代重建技术后图像质量明显改善。25mAs采用idose-4/7重建后图像的肺结节可见度评分优于50mAsFBP重建后的图像（p值<0.001）。【结论】128排螺旋CT实验条件下10mAs可以检出≥5mm的pGGN，结合iDose迭代重建技术后图像质量明显改善，有利于结节边缘情况的显示，为临床方案的制定提供了参考依据。1.3辐射剂量和迭代重建技术对CT测量肺磨玻璃密度结节体积准确性的影响【研究目的】评价辐射剂量和迭代重建技术对测量磨玻璃密度结节（Ground glassnodule, GGN）体积准确性的影响。【材料与方法】应用Philips Brilliance128排螺旋CT在5种不同的管电流水平（管电压120kV不变，管电流200mAs、100mAs、50mAs、25mAs、10mAs）下对含有9个5~12mm的pGGN的仿真胸部体模进行CT扫描，扫描图像采用1mm层厚，骨算法重建，并对重建图像分别采用FBP和迭代重建技术iDose（水平选4）进行重建后传送至工作站，使用Lung Nodule Assessment软件测量模拟GGN的体积。计算每次扫描的有效辐射剂量（ED）。以管电流200mAs+FBP重建为对照组，计算其他各组的相对百分误差（relative percent error, RPE）并进行统计学分析。【结果】10mAs+FBP重建技术条件下，模拟GGN不能被软件成功分离并计算出各自体积。与对照组相比，FBP重建技术下，100mAs、50mAs、25mAs条件下各组的RPE分别为（-0.96±1.28）、(-2.24±2.86)、(-11.97±5.48)，其中25mAs条件下的RPE与其他各组间具有明显的统计学差异（p<0.05），100mAs与50mAs间的RPE无明显统计学差异（p>0.05）。RPE与管电流呈显著负相关关系（r=-0.739，p=0.000）。iDose重建技术下，与对照组相比，100mAs、50mAs、25mAs、10mAs条件下各组的RPE分别为（-0.095±0.39）、(-0.93±0.97)、(-2.18±3.07)、（-3.12±3.85），四组间的RPE具有统计学差异（F=7.605，p<0.001），组间两两比较后发现，10mAs与其余各组间的RPE具有统计学差异（p<0.05），其余各组间的RPE无统计学差异。【结论】不同辐射剂量和重建技术可以对磨玻璃密度结节体积测量产生影响，在肺内磨玻璃密度结节动态观察时，应重视扫描参数和重建技术的一致。与传统的FBP相比，iDose可以使低剂量图像的噪声降低，可以改善图像质量，所获得的磨玻璃密度结节体积误差较小。1.4不同滤波函数和重建技术对低剂量CT肺磨玻璃结节图像质量影响的体模研究【研究目的】评价应用不同滤波函数和重建技术在低剂量胸部CT中评估肺磨玻璃密度结节的可行性。【材料与方法】应用Philips Brilliance128排螺旋CT在4种不同的剂量水平（管电压120kV不变，管电流100mAs、50mAs、30mAs、25mAs）下对含有9个5~12mm的pGGN的仿真胸部体模进行CT扫描，并分别采用滤波反投影（FBP）标准（standard）滤波函数、FBP平滑（smooth）滤波函数和iDose迭代重建技术进行数据重建。对不同剂量水平、不同滤波函数和重建技术下显示的pGGN的CT值、噪声值及其图像质量进行评价，测量并比较每次扫描的有效辐射剂量（ED）。【结果】不同管电流、不同滤波函数和重建技术下所测得的GGN的CT值间均无明显统计学差异（p均>0.05）。各管电流水平下，FBP-standard下所测得的噪声与FBP-smooth、iDose-4/6间具有明显的统计学差异（p值均<0.05）。50mAs、30mAs、25mAs水平下，FBP-smooth下所测得噪声与iDose-4间无明显统计学差异（p值均>0.05）。50mAs条件下，iDose-4下测得的图像噪声接近100mAs+FBP下的噪声。30mAs、25mAs条件下，iDose-6下测得的图像噪声接近100mAs+FBP下的噪声。iDose-4重建的图像质量均较FBP-标准滤波函数、FBP-平滑滤波函数重建的图像质量好，均具有明显统计学差异（p值均<0.05）。【结论】和FBP-标准滤波函数相比，FBP-平滑滤波函数可以降低低剂量图像的噪声，但图像中结节边缘过度平滑，导致部分诊断信息丢失，而iDose迭代重建技术不但可以明显的降低噪声，还可以改善图像质量，并保留了不同结构的对比信息。第二部分迭代重建技术在超低剂量胸部CT的临床应用【研究目的】评价分别采用iDose迭代重建技术（iterative reconstruction，IR）与传统滤波反投影（FBP）重建的低剂量（low-dose diagnostic CT, LDD-CT）和接近胸片剂量的超低剂量胸部CT（ultra-low-dose chest computed tomography, ULD-CT）图像质量的差别。【材料与方法】收集42例行胸部CT平扫的门诊或住院病人，对其同时进行低剂量（120kVp，自动管电流调制技术）和超低剂量（100kVp，10mAs）扫描。图像分别采用FBP，iDose-L4和iDose-L6进行重建。记录各组的CT值、噪声和有效剂量（ED），按1-3分主观评价其肺窗和纵隔窗图像质量，按1-4分评价各组对病变的显示能力。【结果】超低剂量组的有效剂量（ED）为0.215±0.018mSv，低剂量组的有效剂量(ED)为2.6±0.7mSv，两组间的ED具有明显统计学差异(P<0.001)，超低剂量组剂量较低剂量组降低91.73%。和FBP重建的低剂量组相比，超低剂量组采用FBP和iDose-L4重建后图像的CT值明显增高(p<0.05)，而采用iDose-L6重建后的图像CT值则差别不大（p>0.05）。iDose-L4和L6重建后的低剂量组图像的CT值与FBP重建的低剂量组无明显统计学差异（p>0.05）。iDose迭代重建技术可以明显降低SD值，且iDose所选的水平越高，SD值越小。和FBP比较，迭代重建后的图像质量明显改善，迭代重建后的超低剂量组图像可以较好的检出并显示微结节、实性结节、气肿、支扩、实变、纤维灶、胸腔积液、心包积液、胸膜及胸壁病变、纵隔病变的特征，可以检出肺内GGO，但对其边缘及内部特征的显示欠佳，不能充分满足诊断要求。【结论】接近胸片剂量的超低剂量CT扫描技术诊断肺内基本病变时，辐射剂量降低，噪声增加，图像质量也明显下降，诊断信息丢失，但结合迭代重建技术的应用，图像质量明显改善，可满足检出肺内病变的要求，可推荐作为胸部低剂量筛查和随访方案。第三部分迭代重建技术对低管电压胸部CT增强扫描图像质量的影响【研究目的】评价分别采用iDose迭代重建技术（iterative reconstruction，IR）与传统滤波反投影（FBP）重建的常规剂量和低剂量胸部CT增强扫描在体质量指数（BMI）正常（18.5kg/m2≤BMI<25kg/m2）患者中应用的图像质量的差别。【材料与方法】选取80例BMI在正常范围行胸部CT增强扫描的门诊或住院病人，随机分为常规剂量（120KVp）和低剂量（80KVp）组。对常规剂量组采用FBP进行重建（A组），低剂量组则分别采用iDose迭代重建技术（B组）和FBP（C组）进行重建。记录各组的噪声、信噪比（SNR）、对比度噪声比（CNR）和有效剂量（ED），按1-3分评价总体图像质量，并对3组结果进行比较。【结果】低剂量组的有效剂量（ED）较常规剂量组降低71.35%，B组图像噪声较C组小（p<0.05），较C组降低26.74%，且B组SNR、CNR最大，B组总体图像质量与A组无差异（p>0.05）。【结论】和传统滤波反投影FBP比较，迭代重建技术联合低管电压（80KVp）进行胸部CT增强扫描时，可以降低图像噪声，提高图像的SNR和CNR，获得满足诊断要求的图像，此方案可推荐用于胸部血管成像。更多还原

【Abstract】 【Background】Since its introduction, CT has become the major imaging modality for investigatingchest disorders. Compared with chest radiography, CT has greater diagnostic accuracy in arange of clinical situations. This improvement in image quality, however, is associated witha100fold increase in radiation dose compared with conventional radiography. It wasestimated that the number of CT examinations has increased23times, from3million in1980to more than68.7million in2007in the united states. It has estimated that the use ofCT may be responsible for approximately2%of all incident cancer cases in the unitedstates[1]. Therefore, the optimation of CT protocols and reduction in radiation dose havebecome an important focus for research.Providing diagnostic image quality at the lowest radiation dose is an important focus ofCT research in this world. In the NLST[2], a20%decrease in mortality from lung cancerwas observed in the low dose CT group as compared with the radiography group. The rateof positive results was higher with low-dose CT screening than with radiographicscreening. It may be difficult to evaluate GGO nodules by low-dose thin-section helical CT.Li et al[3]reported that32%of lung cancers with solid nodules were missed on CT lungcancer screens performed with low-dose CT. At MDCT, GGO nodules with a CT numberof-650HU or less were difficult to detect at the lower dose settings (21and45mAs)[4].A number of CT dose reduction techniques and researches have been developed withthe goal of preserving image quality while reducing radiation dose. Because the primaryeffect of reduced radiation dose is increased image noise. Most commercial CT systemscurrently use Filtered back projection(FBP) as the standard image reconstruction algorithm.FBP, while fast and fairly robust at routine radiation dose, is prone to image noise andartifacts that result in non-diagnostic images at extremely low doses. As a result, thepotential to reduce the dose while maintaining image quality using FBP is limited. Iterativereconstruction(IR) is now a hot topic to enable the reductions of radiation dose and imagenoise. Numerous IR techniques are available commercially, such as IRIS (iterativereconstruction in image space; Siemens Healthcare, Forchheim, Germany), ASIR(adaptive statistical iterative reconstruction; GE Healthcare, Milwaukee, WI), and iDose(Philips Healthcare) and so on. These techniques can significantly reduce noise andimprove image quality, have a potential to decrease the radiation dose[5-10]. The objective of this work was to investigate the potential value of iterativereconstruction when applying to the low dose chest CT both in a phantom and in patients.Part OneEffect of Iterative reconstruction technique on radiation dose and image qualityfor low dose chest CT: phantom study1.1Impact of low-dose HRCT and iterative reconstruction algorithm(iDose) onthe CT number: a pulmonary phantom study【Objective】 To evaluate the impact of low-dose HRCT and iterative reconstructionalgorithm(iDose) on the CT number using a pulmonary phantom.【Materials andmethods】 HRCT scan were performed on chest phantom with5different tube currents(200mAs、100mAs、50mAs、25mAs、10mAs), scanned with Philips Brilliance128scanner.All data were created using3reconstructions, i.e. filtered back projection (FBP) andmoderate-and high-level iterative reconstructions (iDose-4, iDose-7). Effective dose(ED)was calculated on the basis of dose length product and volumetric CT dose index.Hounsfield unit (HU) and standard deviation (SD) of CT values in different regions of thephantom were recorded.【Results】 No significant differences were found on the CTnumbers obtained using different tube currents or different reconstructions(p>0.05). Andthe SD was increased with the current decrease. Raw data-based iterative reconstructionreduces SD values compared to FBP.【Conclusion】 Low dose HRCT can measure theCT number accurately, And the noise was increased with the current decrease. Rawdata-based iterative reconstruction reduces image noise compared to FBP, The larger theiDose level is, the larger the noise reduction is.1.2Detection of Ground-glass opacities using hybrid iterative reconstruction (iDose)and low-dose HRCT: a phantom study【Objective】 The purpose is to investigate the ability to detect GGO in a chest phantomusing a low-dose HRCT and hybrid IR named iDose.【Materials and methods】 HRCT scan were performed on chest phantom with5different tube currents (200mAs、100mAs、50mAs、25mAs、10mAs), scanned with Philips Brilliance128scanner. All data werecreated using3reconstructions, i.e. filtered back projection (FBP) and moderate-andhigh-level iterative reconstructions (iDose-4, iDose-7). Effective dose(ED) was calculatedon the basis of dose length product and volumetric CT dose index. Standard deviation (SD)of CT values in different regions of the phantom were recorded. Visually assessment ofidentification and image quality of GGNs by two chest radiologists.【Results】Kappavalue of overall image quality were substantial or almost perfect(k>0.4). The identificationof GGN was possible at currents of>10mAs. iDose had better image quality at a lowerradiation dose than images acquired with a conventional FBP reconstruction algorithm.The image quality was significantly lower at50mAs-FBP than at25mAs-iDose（p<0.001).【Conclusion】 GGO nodules of5mm in diameter can be detected with128-slice spiralCT using10mAs at experimental study，The use of hybrid IR techniques enabled improvethe image quality.1.3Impact of radiation dose and iterative reconstruction on the Volumetricmeasurement accuracy of ground-glass opacity nodules【Objective】 To evaluate the impact of the radiation dose and iterative reconstruction onthe volumetric measurement accuracy of ground-glass opacity nodules(GGN) detection.【Material and Methods】 CT scans were performed on chest phantom containing9artificial GGN with5different tube currents (200mAs、100mAs、50mAs、25mAs、10mAs),All data were reconstructed with slice thickness of1.0mm and bone algorithm forreconstruction. The images were reconstructed with regular filtered back projection (FBP)and an iterative reconstruction technique (iDose). Nodule volumes were measured usingLung Analysis software. Effective dose(ED) was calculated. The standard reference usednodule volume measured at200mAs. The relative percent error (RPE) between the CTmeasured volume and the reference volume of GGN were calculated.【Results】 Whentube current was10mAs, volume of GGNs can’t measured using the software. The RPEbetween the CT measured volume and the reference volume of GGN at100mAs,50mAs,25mAs were （-0.96±1.28）、(-2.24±2.86)、(-11.97±5.48) respectively. There werestatistically significant differences between the RPE measured at25mAs and other groups(p<0.05). There was a significant negative correlation between RPE and the tubecurrent(r=-0.739，p=0.000). The RPE of groups reconstructed with iterative reconstruction were （-0.095±0.39）、(-0.93±0.97)、(-2.18±3.07)、（-3.12±3.85）, There were statisticallysignificant differences between the RPE measured at10mAs and other groups (p<0.05).【Conclusion】 The different radiation and reconstruction techniques could have effectson the accuracy of volumetric measurement. Compared with regular FBP, iterativereconstructions enable significant reduction of image noise without loss of diagnosticinformation, thus having the potential to decrease radiation dose and reduce the error ofmeasurement on nodule volume.1.4Effect of kernel and iterative reconstruction on image quality of ground-glassopacities on low-dose CT: A phantom study【Objective】 To evaluate the impact of the kernel and iterative reconstruction on imagequality of ground-glass opacity nodules(GGN) on low-dose chest CT.【Material andMethods】 CT scans were performed on chest phantom containing9artificial GGN with4different tube currents (100mAs、50mAs、30mAs、25mAs), The images werereconstructed with regular filtered back projection (FBP) standard kernel,FBP withsmooth kernel and an iterative reconstruction technique (iDose). After scanning, Wemeasured image noise, CT number Effective dose(ED) and assessed image quality.【Results】 No significant differences were found on the CT numbers obtained usingdifferent tube currents or different reconstructions and kernels(p>0.05). There was astatistically significant difference between FBP with standard kernel and FBP with smoothkernel and iDose-4/6with respect to image noise (p<0.05), Regarding to25mAs,30mAs,50mAs, There were no statistically significant differences between FBPwith smooth kernel and iDose-4(p>0.05). Image noise on50mAs images reconstructedwith iDose-4remained constant compared to100mAs+FBP images. Image noise on25/30mAs images reconstructed with iDose-6remained constant compared to100mAs+FBP images. Image quality was better with iDose than FBP with standard orsmooth kernel.【Conclusion】 From our findings, FBP with smooth kernel can alsoreduce the image noise, but make the lesion more smooth. iDose had better image qualityat a lower radiation dose than images acquired with a conventional FBP reconstructionalgorithm. Part TwoUltra low dose non-enhanced chest CT similar to chest X-ray examination: Potentialvalue of iterative reconstruction【Objective】 The purpose of this study was to assess the diagnostic image quality ofultra-low-dose chest computed tomography (ULD-CT) obtained with a radiation dosecomparable to chest radiography and reconstructed with filtered back projection (FBP) anditerative reconstruction (iDose) in comparison with routine low-dose diagnostic CT(LDD-CT).【Material and Methods】 Unenhanced chest CT images of42patientsacquired with ULD-CT were compared with images obtained with routine LDD-CT in thesame examination. The image noise, CT numbers and effective dose (ED) with eachprotocol were assessed. Normal lung structures and Lesion conspicuity of chest lesions onall CT data sets were assessed on a three-point scale.【Results】 The radiation doseof ULD-CT was0.215±0.018mSv compared with2.6±0.7mSv for LDD-CT(P<0.0001). The CT number of ULD-CT was significantly higher using FBP than thatusing iDose-L4and iDose-L6(p<0.05). Image quality of ULD-CT and LDD-CT increasedsignificantly when using iDose-L4or iDose-L6compared with FBP (P<0.001). ULD-CTreconstructed with iDose enabled to detect the lesions as seen on LDD-CT. However,image quality of ULD-CT reconstructed with iDose was clearly inferior forcharacterisation of ground glass opacities.【Conclusion】 Iterative reconstruction allowsdetection the lesions with ULD-CT with radiation exposure in the range of a posterior toanterior (PA) and lateral chest X-ray.Part ThreeThe effect of iterative reconstruction on image quality of contrast-enhanced chest CTwith low tube voltage settings【Objective】To evaluate the image quality of an iterative reconstruction algorithm(iDose)in contrast-enhanced chest CT with low tube voltage settings in comparison withstandard-dose filtered back projection(FBP) CT in patients with normal body massindex(BMI).【Material and Methods】 Eighty patients with normal body massindex(BMI) were referred for a contrast-enhanced chest CT, The patients were randomlyassigned into groups120-KVp and80-KVp. Standard convolution FBP was used to reconstruct120-KVp (group A) and80-KVp(group C) image sets, and iterativereconstructions(iDose) was used to reconstruct80-KVp (group B) image sets. The meanimage noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and effective dose(ED) with each protocol were assessed. Image quality were graded (scale:1-3) andcompared among3groups.【Results】Radiation dose was71.35%less for the low doseprotocol. Noise was significantly lower in low dose images reconstructed with iDose(Group B)compared to images reconstructed with FBP（Group C). Group B had the highestSNR and CNR. There was no difference in subjective image quality scores between GroupB and A(p>0.05).【Conclusion】 In contrast-enhanced chest CT examinations, Imagesreconstructed with iterative reconstruction had better image quality at a lower radiationdose than images acquired with a conventional FBP reconstruction algorithm.更多还原