(3.3.8)--脑科学与影像新技术.pdf

《(3.3.8)--脑科学与影像新技术.pdf》由会员分享，可在线阅读，更多相关《(3.3.8)--脑科学与影像新技术.pdf（6页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、RESEARCHOpen AccessDiagnosis of brain tumors using dynamiccontrast-enhanced perfusion imaging witha short acquisition timeTakashi Abe1*,Yoshifumi Mizobuchi2,Kohei Nakajima2,Yoichi Otomi1,Saho Irahara1,Yuki Obama1,Mungunkhuyag Majigsuren1,Delgerdalai Khashbat1,Teruyoshi Kageji2,Shinji Nagahiro2and Ma

2、safumi Harada1AbstractThis study sought to determine the diagnostic utility of perfusion parameters derived from dynamic contrast-enhanced(DCE)perfusion MRI with a short acquisition time(approximately 3.5 min)in patients with glioma,brain metastasis,andprimary CNS lymphoma(PCNSL).Twenty-six patients

3、 with 29 lesions(4 low-grade glioma,13 high-grade glioma,7 metastasis,and 5 PCNSL)underwentDCE-MRI in a 3 T scanner.A ROI was placed on the hotspot of each tumor in maps for volume transfer contrast Ktrans,extravascular extracellular volume Ve,and fractional plasma volume Vp.We analyzed differences

4、in parameters betweentumors using the MannWhitney U test.We calculated sensitivity and specificity using receiver operating characteristicsanalysis.Mean Ktransvalues of LGG,HGG,metastasis and PCNSL were 0.034,0.31,0.38,0.44,respectively.Mean Ve values ofeach tumors was 0.036,0.57,0.47,0.96,and mean

5、Vp value of each tumors was 0.070,0.086,0.26,0.17,respectively.Compared with other tumor types,low-grade glioma showed lower Ktrans(P0.01,sensitivity=88%,specificity=100%)and lower Ve(P0.01,sensitivity=96%,specificity=100%).PCNSL showed higher Ve(P 0.01,sensitivity=100%,specificity=88%),but the othe

6、r perfusion parameters overlapped with those of different histology.Kinetic parameters derived from DCE-MRI with short acquisition time provide useful information for thedifferential diagnosis of brain tumors.Keywords:Magnetic resonance imaging(MRI);Dynamic contrast enhanced(DCE)perfusion;Brain tumo

7、r;Two-compartment model analysis;Pharmacokinetic model analysisIntroductionThe differential diagnosis of brain tumor is critical to de-termining optimal therapyand estimating prognosis(DeAngelis 2001).High grade glioma,brain metastasis,and primary central nervous system lymphoma(PCNSL)are common typ

8、es of brain malignancies in adults and cansometimes present similar results on conventional MR im-aging(Ma et al.2010).Dynamic contrast-enhanced(DCE)imaging,which allows for noninvasive evaluation of tumorvascularity,has been widely used to assess the physiologyof brain tumor vascularity(Tofts 1996;

9、Tofts et al.1999).Dynamic acquisition of images during contrast en-hancement allows for the specific descriptive parame-ters related to local microvasculature characteristics tobe calculated.Both relaxivity(T1)-and susceptibility(T2*)-based approaches have demonstrated good poten-tial for measuring

10、the characteristics of tumor vasculature(Quarles et al.2012).Methods to assess changes in tissueT1 following contrast agent injection are commonlytermed DCE-MRI and have been widely performed toassess microvascular permeability(Tofts 1996;Tofts et al.1999).In DCE-MRI,the signal intensity change can

11、bemeasured with sufficient temporal resolution and isrelated to tissue contrast agent concentration.The pharmacokinetic(PK)model introduced by Toftset al.can also be used to calculate volume transfer contrastKtrans,volume of extravascular extracellular space Ve,and*Correspondence:abe.takashitokushim

12、a-u.ac.jp1Department of Radiology,Institute of Health Biosciences,The University ofTokushima Graduate School,3-18-15,Kuramoto-cho,Tokushima City,Tokushima 770-8509,JapanFull list of author information is available at the end of the articlea SpringerOpen Journal 2015 Abe et al.;licensee Springer.This

13、 is an Open Access article distributed under the terms of the Creative CommonsAttribution License(http:/creativecommons.org/licenses/by/4.0),which permits unrestricted use,distribution,and reproductionin any medium,provided the original work is properly credited.Abe et al.SpringerPlus (2015)4:88 DOI

14、 10.1186/s40064-015-0861-6fractional plasma volume Vp(Tofts 1996;Tofts et al.1999).As such,DCE-MRI can provide information on the bloodmicrocirculation of tumors that cannot be acquired fromconventional MRI(Tofts 1996;Tofts et al.1999;Patankaret al.2005;Xyda et al.2012;Sorensen et al.2009;Bisdaset a

15、l.2011;Mills et al.2006).In the brain,previous studieshave used these kinetic parameters to evaluate gliomagrade(Patankar et al.2005),differential diagnosis(Xydaet al.2012),treatment effects in primary brain tumors(Sorensen et al.2009),diagnosing recurrence from radi-ation injury(Bisdas et al.2011)a

16、nd predicting prognosis(Mills et al.2006).DCE data measured with sufficient temporal reso-lution and acquisition time can provide useful results inPK model analysis(Tofts 1996;Larsson et al.2013).Ac-quisition times of over 5 min have been used for thediagnosis of brain tumors in recent years(Bisdas

17、et al.2011;Aref et al.2008;Awasthi et al.2012;Bagher-Ebadian et al.2012;Jia et al.2012)and are recom-mended to maintain reliability(Larsson et al.2013).But due to practical time limitations for an MRI exam-ination,a DCE sequence with a short acquisition timeand high diagnostic performance is require

18、d.AlthoughDCE sequences with short acquisition times result inoverestimated Ktransand underestimated Veand Vp,theresultant error is not so large at acquisition times of 34 min.Further,when using the same DCE protocol fordifferent tumor types,parametric errors occur in thesame direction,so their resu

19、lting distributions could beunchanged and diagnostic utility is preserved.In fact,there have been investigations on brain,head and neck,and breast neoplasms using DCE sequences of less than5 min and the Tofts model(Awasthi et al.2012;ElKhouli et al.2011;Shukla-Dave et al.2012).Addition-ally,for use

20、in clinical practice,reduced operation timeis important and can be achieved with automated post-processing and fixed T1 method(Haacke et al.2007),which uses preselected T1 value as a precontrast T1value(T10)of the target organ to reduce the noise de-rived from T1 map and to reduce total acquisitiont

21、ime.Against this background,the main purpose of thisstudy was to investigate the diagnostic utility of DCE-MRI in diagnosing glioma,metastasis,and PCNSLtumors using a relatively short acquisition time and to re-duce operation time using automated post-processing anda fixed T1 method.Materials and me

22、thodsThis study was approved by the university hospital ofTokushima clinical trial center for developmental thera-peutics,and informed consent was obtained from allpatients prior to enrollment.PatientsFifty-two consecutive patients who underwent contrast-enhanced MRI for diagnosing suspected brain t

23、umor be-tween December 2012 and December 2013 were eligiblefor this study.From this group,those with glioma,metasta-sis,and PCNSL were included in the analysis.Diagnosiswas made histologically or clinicoradiologically.Clinicora-diological diagnosis was made by consensus of two experi-enced neuroradi

24、ologists.Metastases were defined as newlyemerging nodules in cancer patients;and recurrence of gli-oma were defined as a steady increase in contrast enhancedT1-weighted images.MRI follow-up was performed at2-month intervals or sooner.Twenty-six patients(17 menand 9 women;mean age 61 years,age range

25、3584,years;29 lesions)were included in the analysis.The time fromexamination to diagnosis was recorded.We dividedtumors into four groups:low grade glioma(LGG),highgrade glioma(HGG),metastasis,and PCNSL.Table 1summarizes the patient information.Imaging protocolExaminations were performed with a 3 T M

26、R scanner(Discovery 750,GE Healthcare,Milwaukee,WI)usinga standard eight-channel head coil.Pre-contrast T1-weighted images,T2-weighted images,diffusion-weightedimages,arterial spin labeling images,and MR spectros-copy were acquired.Subsequently,DCE-MRI and post-contrast T1-weighted images were acqui

27、red with contrastagent(Gd-DTPA,0.1 mmol/kg;Magnevist,Bayer Health-Care,Berlin,Germany).We performed 3D T2*-weightedangiography and DTI in selected patients.DCE-MRI was acquired using a 3D fast spoiled gradi-ent echo sequence with TR=4.4 ms,TE=1 ms,flipangle=12,field of view=300 210 mm,matrix=128 90,

28、slice thickness=8 mm,and number of slices=16,consisting of 64 phases with a temporal spacing of 3.3 s.We chose TR,TE,and flip angle in accordance withthose of an ordinary spoiled gradient echo sequence.Table 1 A summary of patient informationNo.of cases(male)Mean age,years(range)PathologyLGG4(1)53.3

29、(3577)3 oligodendrogliomas,1diffuse astrocytomaHGG13(9)59.2(3484)1 anaplastic astrocytoma,1anaplastic oligodendroglioma,1 gliosarcoma,8 glioblastomas,2 recurrent high grade gliomaMetastasis6(4)64.3(4877)2 lung cancers*,2 breastcancers*,1 gastric cancer,1colon cancer*Primary CNSlymphoma3(3)69.0(5578)

30、3 diffuse large B-celllymphomas*:Three cases were diagnosed clinically(1 lung cancer,1 breast cancer andcolon cancer).The others were diagnosed pathologically.Abe et al.SpringerPlus (2015)4:88 Page 2 of 6Total scan time was 3 min and 31 s.A gadolinium-contrast agent was injected with a power injecto

31、r(Medrad,Indianola,PA)at a rate of 2.5 ml/s after twocycles of dynamic scan.Immediately afterward,20 ml ofsaline was injected at the same rate.Image analysisAll imaging data were transferred from the scanner to aworkstation(Advantage Workstation 4.6,GE MedicalSystems,Milwaukee,WI).We analyzed DCE da

32、ta usingthe Tofts model implemented in the commercially avail-able software GenIQ(GE Medical Systems,Milwaukee,WI).We used 3-dimensional,rigid motion correction,which conduct a combination of rotational movementand translation in all cases.Referring to the temporalchanges of the signal intensity of

33、all voxel,the softwareautomatically extracts the pixels that considered to ar-teries and veins.In the analysis,we used fixed T1method(Haacke et al.2007)and default T10 value(T10=1000 ms).Data analysisWe selected enhanced lesions with a minor axis of morethan twice the slice thickness(16 mm)and inclu

34、de themin the following analysis.In each map,we set an ROI ofapproximately 100 mm2at the hot spot of the tumor.We measured average Ktrans,Ve,and Vpin each tumor.We then assessed the correlation of PK model parame-ters with different parametric maps and tumor histology.Statistical analysisWe first ca

35、lculated correlations between the DCE parame-ters Ktrans,Ve,and Vpusing Spearmans rank correlationcoefficient.We then assessed the correlation betweenDCEparametersandtumorhistology.Resultsareexpressed as meanstandard deviation.Statistical differ-ence between tumors was determined using the MannWhitn

36、ey U test.A P value of less than 0.05 was consideredstatistically significant.Finally,we assessed the utility of Ktrans,Ve,and Vpindiagnosis of the brain tumors.From the results of thisanalysis,we selected a group of tumors with distinct PKparameters.We performed receiver operating characteris-tics(

37、ROC)curve analysis for selected tumors to evaluatethe optimal cutoff value,sensitivity,and specificity.All statistical analysis was performed using ExcelStatistics 2012(Social Survey Research Information Co.,Ltd.,Tokyo,Japan)with Excel 2010(Microsoft Co.,Redmond,WA).ResultsThe time from examination

38、to diagnosis was 5 days(3 and12 days:25thand 75thpercentiles).Data transfer andpost-processing took approximately 1012 min.Contrast-enhanced T1-weighted imaging for a repre-sentative glioblastoma case(Figure 1)showed increasedKtrans,Ve,and Vp.There was an intermediate correlationbetween Ktransand Ve

39、(R2=0.41),while Vpshowed rela-tively weak correlations with Ktransand Ve(R2=0.17 and0.18,respectively)LGG showed lower Ktransand Vethan the other malig-nant tumor types(P0.01).Among the other tumortypes,Ktransvalues overlapped.Lymphoma showed ex-tremely high Ve(P0.01),but Vefor HGG and metastasisove

40、rlapped.No statistical differences were found for Vp(Figure 2).Area under the ROC curve for differentiating LGGand PCNSL was highest for Ve(LGG:0.97,PCNSL:0.95).A cutoff value of Ktrans=0.0848 for diagnosis ofLGG provided the best good combination of sensitivityand specificity(0.88 and 1.0,respectiv

41、ely).A cutoff valueof Ve=0.18 for diagnosis of LGG provided the best goodcombination of sensitivity and specificity(0.96 and 1.0,respectively).A cutoff value of Ve=0.912 for diagnosisof PCNSL provided the best combination of sensitivityand specificity(1.00 and 0.88,respectively).DiscussionThe result

42、s of this study indicate that kinetic parametersacquired from DCE perfusion study with short acquisi-tion time supplement conventional imaging in predict-ing tumor histology.This method takes only about15 min(scan time of 3.5 min plus post-processing timeof 1012 min)and is not subject to operator-de

43、pendentbias,making it highly feasible in clinical settings.Ktransand Vedemonstrated modest correlation witheach other.Vewas the most useful parameter in diag-nosing LGG and PCNSL,while Ktranswas effective indifferentiating LGG from the other tumors.Vpfailed toprove useful in differentiating brain tu

44、mor types in thisstudy group.Although the utility of the two-compartment PKmodel methods in brain tumor diagnosis has been re-ported(Patankar et al.2005;Xyda et al.2012;Sorensenet al.2009;Bisdas et al.2011;Aref et al.2008;Awasthiet al.2012;Bagher-Ebadian et al.2012;Jia et al.2012),ithas not extended

45、 to clinical practice.One of the reasonsfor this is the long acquisition time.In the present study,we confirmed that the diagnostic performance of DCEanalysis using a short acquisition time is comparable tothat of methods in previous studies.We believe,there-fore,that this method provides new and us

46、eful perform-ance improvements for tumor diagnosis.Another method,first-pass pharmacokinetic model(FPPM)analysis,can be performed from DCE analysisdata with an ultra-short acquisition time of about 1 min(Li et al.2000).Ktransand Vpcan also be calculated withthis method and are comparable to the data

47、 obtainedAbe et al.SpringerPlus (2015)4:88 Page 3 of 6Figure 1 Magnetic resonance imaging of a glioblastoma.A,Axial post-contrast T1-weighted image shows a ringed enhanced lesion in theleft thalamus and subtle enhancement in the right thalamus.B,C,and D,Three kinetic parametric maps show increased v

48、ascular permeability(B;Ktransmap),leakage space(C;Vemap),and plasma volume(D;Vpmap)corresponding to the enhanced area on the contrast-enhanced MRI.Figure 2 Scatter plot(meanstandard deviation)shows 3 kinetic parameters for 4 brain tumor types.A:Ktrans,B:Ve,C:Vp.LGG:low gradeglioma,HGG:high grade gli

49、oma,PCNSL:primary CNS lymphoma,*:significant difference(P0.05,MannWhitney U test)between tumor groups.Abe et al.SpringerPlus (2015)4:88 Page 4 of 6from conventional PK model analysis(Harrer et al.2004).Since,in the FPPM method,tracer concentrationin arterial blood plasma is assumed to be much larger

50、than that in the extravascular extracellular component,Vecannot be calculated(Li et al.2000).Although con-ventional PK model analysis requires a longer acquisi-tion time than the FPPM method,we believe the utilityof Vein the diagnosis of brain tumors justifies the longeracquisition time of conventio