【作者】 单飞；

【导师】 曾蒙苏； 张志勇； 葛棣； 胡洁； 杨姗；

【作者基本信息】 复旦大学 ， 影像医学与核医学， 2009， 博士

【摘要】 第一部分肺癌CT灌注首过期时间分辨率优化研究目的:优化肺癌首过期CT灌注（CT perfusion,CTP）的时间分辨率（temporalresolution,TR）,降低辐射剂量。材料与方法:病理证实的肺癌患者60例,行64层螺旋CT灌注检查,采用Cine模式、图像间cine时间（即TR）0.5s、屏气30s扫描。以1幅图像代表整个肿瘤,通过GE Perfusion 3获得BF、BV、MTT、PS参数。通过剔除TR0.5s序列部分图像,得到TR分别为1～4s的6组数据。以不同TR为固定变量,肿瘤大小为协变量行CTP参数的随机区组协方差分析,以TR 0.5s组为参照,行两两比较。结果:不同TR组的BV与PS值无统计学差异,P均＞0.05,BF与MTT值有统计学差异,P均＜0.05。与TR 0.5组相比,TR 1、1.5、2s组的BF及MTT值无统计学差异;TR 2.5、3及4s组的BF及MTT值有统计学差异,P均＜0.05。结论:TR 1.5s,可保证CTVP参数的准确性,降低辐射剂量。第二部分临床病理因素对NSCLC的CT容积灌注的影响分析目的:了解NSCLC去卷积法CT容积灌注（CTVP）的临床病理影响因素。材料与方法:收集在本院行64-SCT首过期CTVP检查,且经病理或细胞学证实的NSCLC 95例。CTVP检查采用Cine或轴扫模式,50ml造影剂跟随30ml生理盐水以5ml/s速率经肘前静脉注射。使用GE Perfusion 3软件包计算CTVP参数BF、BV、MTT及PS值。采用Pearson相关分析NSCLC各CTVP参数与肿瘤大小的相关性。采用完全随机的t检验或Mann-Whitney非参数检验对比中央型与周围型肺癌两组间各CTVP参数有无差异;采用完全随机的t检验对比坏死组与无坏死组,腺癌组与鳞癌组各CTVP参数有无差别。采用One-way ANOVA对比不同临床分期、淋巴结转移分期的NSCLC各CTVP参数有无差异。结果:96例NSCLC的BF及PS值与肿瘤大小呈负相关;MTT值与肿瘤大小呈正相关,P均＜0.05;43例无坏死肺癌BV值与肿瘤大小呈正相关,P＜0.05。中央型NSCLC的BF及PS值低于周围型NSCLC,P均＜0.05。中央型与周围型NSCLC的BV值及MTT值差异无统计学意义。无坏死NSCLC的BF、BV、PS值高于、MTT值低于坏死NSCLC,P均＜0.05。肺腺癌的BF及PS值高于肺鳞癌,P均＜0.05;BV及MTT值无统计学差异。不同临床分期的NSCLC灌注参数均无统计学差异。不同淋巴结分期的NSCLC灌注参数均无统计学差异。结论:肿瘤大小、坏死、部位和病理类型是影响NSCLC CTVP的因素。在运用CTVP诊断SPN和监测抗血管生成类药物治疗晚期NSCLC时需要考虑这些因素的影响。第三部分CT容积灌注在SPN诊断中的应用价值第一节CT容积灌注的SPN诊断价值及与血管参数的相关性研究目的:评价CTVP在SPN诊断和鉴别诊断中的价值,探讨CTVP与微血管密度（MVD）及有腔血管数目（LVN）、有腔血管面积（LVA）及有腔血管周长（LVP）的相关性。材料和方法;收集经手术或临床治疗随访证实,并行首过期CTVP检查的SPN 65例。使用GE Perfusion 3软件获得CTVP参数BF、BV、MTT、PS值。65例SPN中38例行CD34及α-平滑肌（SMA）免疫组织化学染色。MVD计数时,任何被CD34抗体棕黄染的、与邻近结构明确分离的单个细胞或细胞簇,即使未显示管状结构均计为一个血管;任何有条状、裂隙状或点状管腔的被SMA抗体棕黄染的血管定义为有腔/成熟血管,并通过软件测量LVN、LVA及LVP。统计分析分两部分。第一部分:先对比恶性与良性结节组BF、BV、MTT、PS均数有无差异,再作鉴别恶性和良性结节组的ROC曲线,并比较诊断正确性有无差异,后取整数鉴别参数值。最后根据不同CTVP参数的敏感性、特异性,进行两项参数的联合试验。第二部分:比较38例行免疫组化染色的恶性与良性结节的CTVP及血管参数有无差异。采用Pearson或Spearman相关分析各CTVP参数与MVD、各有腔血管参数的相关性。结果:恶性与良性结节组的BF、BV、PS值均数间有统计学差异,P均＜0.001;MTT值无统计学差异。BF、BV、PS值可鉴别SPN良恶性,AUC_ROC分别为0.844、0.854、0.831,P均＜0.001;三条ROC曲线间诊断效能无统计学差异。BF=50 ml/min/100g时,敏感性93%、特异性59%;BV=3 ml/100g时,敏感性86%、特异性68%;PS=10ml/min/100g时,敏感性86%、特异性77%。BF+PS及BV+PS的序列实验联合特异性均为82%。行免疫组化染色的38例SPN中,恶性与良性结节组的BF、BV、PS、LVA、LVP及LVN差异有统计学意义,P均＜0.05;MTT及MVD差异无统计学意义。BF、MTT及PS值与MVD无相关性;BV值与MVD有相关性,r=0.483,P＜0.05;BF值与LVA、LVP、LVN间均有相关性,r值分别为0.547、0.507、0.404,P均＜0.05;BV值与LVA、LVP间有相关性,r值分别为0.714、0.504,P均＜0.05,与LVN间无相关性,r=0.260,P＞0.05;PS值与LVA、LVP、LVN间均有相关性,r分别0.507、0.541、0.473,P均＜0.05;MTT值与LVA、LVP及LVN间均无相关性,P均＞0.05。结论:CTVP可用于SPN诊断。CTVP更多地反应有腔/成熟血管,与有腔血管参数的相关性较MVD好。CTVP应替代MVD作为无创性评价血管生成、血管反应的方法。第二节CT容积灌注诊断SPN的准确性和一致性研究目的:探讨CTVP结合常规胸部CT诊断SPN的准确性和一致性。材料与方法:收集经手术或临床治疗随访证实,并行首过期CTVP检查的SPN 65例。CTVP数据使用GE Perfusion 3进行处理。由两位低年资和两位高年资医生盲法先阅常规胸部CT图像,做出诊断（方法A）;再阅常规胸部CT及CTVP资料,作出诊断（方法B）。采用ROC曲线比较每一位医生及不同医生间诊断的正确性有无统计学差异。通过组内相关系数（intra-class correlation coefficient,ICC）评价四位医生的SPN诊断的一致性。结果:两位低年资医生诊断方法A的AUC_ROC为0.748、0.765,诊断方法B的AUC_ROC为0.967、0.950;两位高年资医生诊断方法A的AUC_ROC为0.894、0.901;诊断方法B的AUC_ROC为0.978、0.967。方法A低年资医生间、高年资医生间AUC_ROC差别无统计学意义;高低年资医生间AUC_ROC差别有统计学意义,P均＜0.05。四位医生方法B的AUC_ROC均大于方法A,P均＜0.05。四位医生间方法B的AUC_ROC差异均无统计学意义,P均＜0.05。四位医生方法A的ICC=0.612;方法B的ICC=0.854。结论:联合运用常规CT及CTVP较单独运用常规CT可提高放射科医生诊断SPN正确性及一致性。第四部分肺癌恩度联合化疗CT容积灌注改变的研究第一节NSCLC的64层螺旋CT容积灌注可重复性研究目的:探讨NSCLC CT容积灌注（CTVP）的可重复性。材料与方法:收集在本院行首过期同层动态64-SCT容积灌注检查（8×5mm层厚）,且经病理证实的NSCLC患者60例,其中14例在24h内行第二次CTVP检查。60例患者中,肿瘤最大径≤3cm及＞3cm各30例。14例行两次CTVP检查的患者中,肿瘤最大径≤3cm及＞3cm各7例。通过GE perfusion 3软件包,两名放射科医生盲法计算60例患者BF、BV、MTT及PS值;一名放射科医生计算14例行两次检查的参数值。采用组内相关系数（ICC）及Bland-Altman法评价观察者一致性及两次CTVP检查的可重复性。结果:NSCLC最大径≤3cm组的CTVP参数观察者ICC均＞0.75;BF、BV、MTT值的观察者内ICC及RC均大于观察者间ICC及RC。NSCLC最大径＞3cm组的CTVP参数观察者内ICC均＞0.75;BF及BV值的观察者间ICC均＞0.75,MTT及PS值的ICC均＜0.75;BV、MTT及PS值观察者内ICC均大于观察者间ICC,CTVP参数观察者内RC均大于观察者间RC。两次CTVP重复检查,NSCLC最大径≤3cm组的BF、BV及PS值可重复性ICC均＞0.75,RC依次为56%、45%、78%;NSCLC最大径＞3cm组的BF、BV及PS值可重复性ICC均＞0.75,RC依次为46%、30%、33%。最大径≤3cm组BF值的RC 95%变化区间为-39～53%,BV值为-29～62%,MTT值为-83～145%,PS值为-57～98%;最大径＞3cm组BF值95%变化区间为-48～45%,BV值为-33～26%MTT值为-54～64%,PS值为-18～48%。结论:NSCLC的CTVP观察者内一致性总体好于观察者间一致性。肿瘤最大径＞3cm的NSCLC可重复性优于最大径≤3cm的NSCLC。BF及BV值适合监测抗血管生成药物治疗晚期NSCLC疗效。第二节晚期NSCLC恩度联合化疗CT容积灌注早期改变的初步研究目的:了解重组人血管内皮抑素（恩度）治疗晚期NSCLC后的早期CTVP改变。材料与方法:晚期初治NSCLC患者15例分为恩度联合化疗和单纯化疗二组,其中恩度联合化疗组9例,单纯化疗组6例。二组均在治疗前和第一周期治疗后1～10天内行二次30秒首过期64-SCT容积灌注检查（Cine模式,8×5mm层厚）。一名放射科医生通过GE perfusion 3软件包计算CTVP参数BF、BV、MTT及PS值,并计算每次靶灶的RECIST-长径和。以Wilcoxon非参数检验了解治疗前、后两组的CTVP参数值及长径和有无差异。以确切概率法卡方检验了解恩度联合化疗组出现灌注变化的比率是否高于单纯化疗组。结果:恩度联合化疗组治疗后BF值高于治疗前,P=0.038;余CTVP参数及长径和治疗前后差异无统计学意义。单纯化疗组治疗前后CTVP参数差异均无统计学意义;治疗后长径和小于治疗前,P=0.028。恩度联合化疗组治疗后CTVP有变化的比率高于单纯化疗组,P=0.047。结论:晚期NSCLC恩度联合化疗后,初步发现早期表现为CT灌注增高。CTVP可发现晚期NSCLC抗血管生成治疗后的血液动力学变化。更多还原

【Abstract】 Part oneFirst pass CT perfusion with lung cancers:usage of different temporal resolutionsPurpose:To find a good temporal resolution（TR） and reduce radiation dose in the CT perfusion（CTP） with lung cancers.Material and methods:Sixty patients with pathologically proved lung cancers were enrolled in a 30s dynamic first pass CTP （cine mode,TR of 0.5s）.Parametric maps of one central slice for BF,BV,MTT and PS were calculated for TR datasets of 0.5,1,1.5,2,2.5,3 and 4s using GE’s Perfusion 3 software package.The analysis of covariance（AOCV） was used to judge whether the differences were found in values of BF,BV,MTT and PS with different TR datasets.In AOCV,the dependent variable was BF,BV,MTT and PS values, fixed factor was different groups of TR,and the covariate was the group of tumor size. Finally,each different TR datasets was compared with the reference dataset（TR of 0.5s） using Sidak test.Results:Statistically significant differences were found in means of values of BF and MTT with different TR datasets（P＜0.001）.No statistically significant differences were found in means of values of BV and PS with different TR datasets（P＞0.05）.Compared with TR dataset of 0.5s,no statistically significant differences were found in means of BF and MTT with TR datasets of 1s, 1.5s,2s（P＞0.05）,while statistically significant differences were found in means of BF and MTT with TR datasets of 2.5,3s,4s（P＜0.05）.Conclusions:When an axial scan with no table feed mode is used,it is recommended that TR of 1.5s should be taken and the radiation dose will be greatly reduced,and without a loss in quantitative accuracy in the CTP with lung cancers. Part twoVolume-based perfusion of NSCLC:evaluation of clinical and pathological factorsPurpose:To investigate the clinical and pathological factors which could be influence the deconvolution-based CT perfusion imaging with NSCLC at 64-slice CT.Material and methods:Ninety five patients with pathologically or cytologically proved NSCLC underwent the first pass CT volume-based perfusion（CTVP） examination at 64-slice CT（cine or axial scan mode with no table feed after intravenous injection of 50 ml contrast media followed by a saline flush of 30 ml at injection rate of 5 ml/s）.BF,BV,MTT,and PS were determined with GE perfusion 3 software package.Pearson correlation coefficient was used to correlate the CTVP parameters with the tumor size.Student t statistics or Mann-Whitney nonparametric statistics were used to compare the differences of CTVP parameters between central and peripheral lung cancers.Student t statistics were applied to compare the differences of CTVP parameters between necrotic and no necrotic NSCLC,and between lung adenocarcinomas and squamous cell carcinomas.One-way ANOVA was employed to compare the differences of CTVP parameters among clinical stags,and among stages of lymph nodular metastases.Results:In 96 cases with NSCLC,BF and PS values were negatively,and MTT values were positively correlated with tumor size（r=-0.270,-0.329 and 0.379,respectively,and P＜0.05 for all）.In 43 patients of no necrotic NSCLC,BV values was positively correlated with tumor size （r=0.502,P＜0.05）.BF and PS values with central NSCLC were lower than these with peripheral NSCLC（P＜0.05 for both）,not MTT and BV values.BF、BV、PS values with no necrotic NSCLC were higher than these with necrotic NSCLC,not MTT values.BF and PS values with lung adenocarcinomas were higher than these of lung squamous cell carcinomas（P＜0.05 for both）,not MTT and PS values.No statistically differences were found among BF values with NSCLC of different clinical stages,and so did BV,MTT,and PS values（P＞0.05 for all）.No statistically differences were found among BF values with NSCLC of different stages of lymph nodular metastases, and so did BV,MTT,and PS values（P＞0.05 for all）.Conclusions:Tumor size, necrosis,tumor location and pathological types could influence the CTVP parameters with NSCLC,and should be addressed for clinical application in diagnosis of SPN and antiangiogenesis therapeutic monitoring. Part threeCT volume-based perfusion with SPNs:clinical evaluation Section one SPNs:CT volume-based perfusion differences in malignant and benign lesions and correlations with vascular parametersPurpose:To investigate the diagnostic value of SPNs with CTVP and the relationships between CTVP and microvascular density（MVD） and luminal vascular number（LVN） and luminal vascular area（LVA） and luminal vascular perimeter （LVP）.Material and methods:Sixty-five patients with proved SPNs were enrolled in a dynamic first pass CTVP.BF,BV,MTT,and PS were determined by GE Perfusion 3 software package.Thirty-eight patients were underwent immunohistochemical staining with CD34 andα-smooth muscle action（SMA）.When MVDs were counted, any CD34 highlighted single endothelial cell or cell cluster that was clearly separate from adjacent tissue elements was counted as a single countable microvessel.For LVN,LVA,and LVP,only SMA highlighted with a discernible lumen could be defined as a luminal/mature vessel,and measured by imaging analysis software.The statics was undergone by two parts.In the first part,the means of four perfusion parameters between malignant and benign nodules were compared.Then,ROC curve for perfusion parameters between malignant and benign nodules were used,and the diagnostic accuracy were compared through the areas under ROC curves,and the integral cut-off were taken.Finally,according to the sensitivities and specificities of different perfusion parameters,the combined tests were done.In the second part, perfusion and vascular parameters between malignant and benign nodules were compared,which were 38 nodules underwent immunohistochemical staining.Pearson or Spearman correlation coefficients were used to investigate relationships between perfusion and vascular parameters.Results:BF,BV and PS values of malignant nodules were higher than those of benign nodules.Statistically significant differences were found between BF and BV and PS values of malignant nodules with those of benign nodules,respectively（P＜0.001 for all）,not MTT values.BF,BV and PS values could be used to judge whether a SPN was benign or malignant,the AUCRoc were 0.844 and 0.854 and 0.831,respectively（P＜0.001 for all）.No statistically significant differences were found in the diagnostic accuracy with three ROC curves. When BF=50 ml/min/100g was used,the sensitivity was 93%,the specificity was 59%.When BV=3 ml/100g was used,the sensitivity was 86%,the specificity was 68%.When PS=10 ml/min/100g was used,the sensitivity was 86%,the specificity was 77%.The combined specificity of the serial test combined BF with PS value and combined BV with PS value were all 82%.In those patients having SPNs stained with CD34 and SMA,statistically significant differences were found in BF,BV,PS,LVA, LVP,LVN of malignant nodules and those of benign nodules,respectively（P＜0.05 for all）;not MTT values and MVD,respectively.BV value were positively correlated with MVD（r=0.483,P＜0.05）,but not the values of BF and MTT and PS,respectively. Significant correlations were observed between BF value and LVA and LVP and LVN（r=0.547 and r=0.507 and r=0.404,respectively,P＜0.05 for all）.Significant correlations were observed between BV value and LVA and LVP（r=0.714 and r=0.504,respectively,P＜0.05 for both）,but had not LVN.Significant correlations were observed between PS value and LVA and LVP and LVN（r=0.507 and r=0.541 and r=0.473,respectively,P＜0.05 for all）,Significant correlations were not observed between MTT value and LVA and LVP and LVN,respectively.Conclusions:CTVP could be used to the diagnose of SPNs.CTVP could reflect luminal/mature vessels more,and had better correlations between luminal vascular parameters than MVD. CTVP should take the place of MVD as a non-invasive method to investigate the state ofangiogenesis or vessel reflections.Section two SPNs:diagnostic accuracy and agreement with CT volume-based perfusion and routine thorax CTPurpose:To investigate whether CT volume-based perfusion（CTVP） combined with routine thorax CT examination could improve the accuracy and agreement for the diagnosis of SPNs by radiologists.Material and methods:Sixty-five patients with pathological and clinical proved SPNs were enrolled in a dynamic first-pass CTVP study.As a part of the study,routine thorax CT examination including non-enhanced and enhanced scan in delayed phase were involved.Perfusion images were preceded by GE Perfusion 3 software package.Two junior and two senior radiologists made a diagnosis by means of those routine thorax CT datasets（method A） blandly.After a month,they made a diagnosis based on CTVP and routine thorax CT datasets（method B） blandly.ROC curves were used to analyses every doctor’s diagnosis made by method A and B.ROC curves before and after CTVP used were compared in every doctor.ROC curves for method A and B with different doctor were compared, respectively.Intra-class correlation coefficients（ICC） were used to investigate the agreements for diagnosis made by four radiologists before and after CTVP used, respectively.Results:With method A,the AUC_ROC in two junior and two senior radiologists were 0.748,0.765,0.894 and 0.901,respectively.With method B,those were 0.967,0.950,0.978 and 0.967,respectively.By AUC_ROC analysis,with method A,no significant statistical differences were found in accuracy of junior or senior radiologists,while significant statistical differences were found in accuracy between junior and senior radiologists（P<0.05 for all）.After CTVP datasets were used,no significant statistical differences were found in accuracy of every radiologist.While method B were compared to method A,significant statistical differences were found in accuracy of every radiologists,respectively（P<0.001 for all）.With method A,ICC for four radiologists were 0.612.After CTVP datasets were used,the ICC was increased to 0.854.Conclusions:Compared to routine thorax CT used singly,usage of the datasets of CTVP and routine thorax CT could improve the accuracy and agreement for radiologists’ diagnosis of SPNs.Diagnosis of SPNs with CTVP plus routine thorax CT should be addressed in clinical application. Part fourLung cancer CT volume-based perfusion:evaluation of combining endostar with chemotherapySection one NSCLC perfusion at 64-Slice CT:reproducibility of volume-based quantitative measurementPurpose:To know the reproducibility of volume-based quantitative measurement of NSCLC perfusion using 64-slice CT.Materials and methods:Sixty patients with pathological-proved NSCLC were enrolled in this dynamic first pass volume-based CT perfusion study（8×5 mm collimation）,and 14 patients of them were underwent the second scan within 24h.Those 60 patients were divided into≤3cm and＞3cm groups according to the longest diameters,and each group had 30 patients.Fourteen patients who were underwent twice CTVP scans were also classified to≤3cm and＞3cm groups according to the same standard,and each group had 7 patients. Using GE perfusion 3 software package,two independent radiologists determined the values of BF、BV、MTT and PS.A radiologist determined the perfusion of 14 patients who underwent twice scans.Intraclass correlation coefficient（ICC） and Bland-Altman statistics were used to evaluate the agreement of observer and the reproducibility of CTVP imaging.Results:In NSCLC（≤3 cm）,the agreements of observer with CTVP parameters were good（ICC＞0.75 for all）.In the group,with BF, BV,and MTT values,the values of ICC and RC for the agreement of intra-observer were higher than those for the inter-observer.In NSCLC（＞3 cm）,the agreements of intra-observer with CTVP parameters were good（ICC＞0.75 for all）.In that group, with BF and BV values,the agreements of inter-observer were good（ICC＞0.75 for both）,while with MTT and PS values,which were moderate or bad（ICC＜0.75 for both）.Also in all perfusion parameters of that group,the values of ICC and RC for the agreement of intra-observer were higher than those for the inter-observer,except ICC with BF value.In both two groups of NSCLC,the reproducibility with BF,BV,and PS values were good（ICC＞0.75 for all）.For NSCLC（≤3 cm）,RC values with BF, BV,MTT and PS values were 56%,45%,114%,and 78%,respectively,and the 95% change intervals of RC were -39～53%,-29～62%,-83～145%,and -57～98%, respectively.For NSCLC（＞3 cm）,those values were 46%,30%,59%,and 33%, respectively,and the 95%change intervals of RC were- 48～45%,-33～26%,-54～64%, and -18～48%.Conclusions:There is greater inter- than intraobserver agreement for CTVP measurements of NSCLC,generally.There is greater reproducibility of tumor size＞3cm than tumor size≤3cm.BF and BV could be addressed for reliable clinical application in antiangiogenesis therapeutic monitoring with NSCLC patients.Section two Usage of CT volume-based perfusion for combining endostar with chemotherapy in NSCLC patients:initial experiencePurpose:The aim of this study was to know whether CT volume-based perfusion （CTVP） could early display the therapeutic activity for combing Rh-endostatin （endostar） with chemotherapy in advanced NSCLC patients.Materials and methods: Fifteen patients with initial treated advanced NSCLC were divided into two groups. Nine patients in the trial group were treated with endostar plus chemotherapy,and 6 patients in the control group were treated with single chemotherapy.Patients in two groups were all underwent 30s first pass CTVP（Cine mode,8×5 mm slice thickness） before and after treatment of first cycle.A radiologist used GE perfusion 3 software to get tumor BF,BV,MTT,and PS.The sums of longest diameters of the appointed target lesions in the transversal plane for response evaluation criteria in solid tumors （RECIST） were also calculated.Wilcoxon nonparametric statistic was used to compare BF,BV,MTT,PS,and sum of longest diameters between two groups.Fish’s exact test was employed to compare the ratio of CTVP difference of first cycle to baseline between two groups.Results:In the trial group,BF values after first cycle were higher than the baseline（P=0.038）,not other CTVP parameters and the sum of longest diameters.In the control group,the sum of longest diameters after treatments were smaller than baseline（P=0.028）,not all perfusion parameters.The ratio of CTVP difference of first cycle to baseline in the trial group was higher than these in the control group（P=0.047）.Conclusions:In the treatment of combing endostar with chemotherapy in advanced NSCLC patients,tumor perfusion would be increase early. CTVP is feasible in showing changes ofperfusion induced by antiangiogenic therapy.更多还原