clinical utility of dual-energy ct in the evaluation of ... 3 - new... · gle-energy ct (p.67)....

Clinical Utility of Dual-Energy CTin the Evaluation of SolitaryPulmonary Nodules: InitialExperience1

Eun Jin Chae, MDJae-Woo Song, MD, PhDJoon Beom Seo, MD, PhDBernhard Krauss, PhDYu Mi Jang, MDKoun-Sik Song, MD, PhD

Purpose: To determine the clinical utility of dual-energy computedtomography (CT) in evaluating solitary pulmonary nodules(SPNs).

Materials andMethods:

This study was approved by the institutional review board,and informed consent was obtained. CT scans were ob-tained before and 3 minutes after contrast material injec-tion in 49 patients (26 men, 23 women; mean age, 60.39years � 12.24 [standard deviation]) by using a scannerwith a dual-energy technique. Image sets that includednonenhanced weighted average, enhanced weighted aver-age, virtual nonenhanced, and iodine-enhanced imageswere reconstructed. CT numbers of SPNs on virtual non-enhanced and nonenhanced weighted average imageswere compared, and CT numbers on iodine-enhanced im-age and the degree of enhancement were compared. Diag-nostic accuracy for malignancy by using CT number oniodine-enhanced image and the degree of enhancementwere compared. On the virtual nonenhanced image, thenumber and size of calcifications were compared withthose on the nonenhanced weighted average image. Radi-ation dose was compared with that of single-energy CT.

Results: CT numbers on virtual nonenhanced and nonenhancedweighted average images and CT numbers on the iodine-enhanced image and the degree of enhancement showedgood agreements (intraclass correlation coefficients: 0.83and 0.91, respectively). Diagnostic accuracy for malig-nancy by using CT numbers on iodine-enhanced image wascomparable to that by using the degree of enhancement(sensitivity, 92% and 72%; specificity, 70% and 70%; ac-curacy, 82.2% and 71.1%, respectively). On virtual non-enhanced image, 85.0% (17 of 20) of calcifications in theSPN and 97.8% (44 of 45) of calcifications in the lymphnodes were detected, and the apparent sizes were smallerthan those on the nonenhanced weighted average image.Radiation dose (average dose-length product, 240.77mGy � cm) was not significantly different from that of sin-gle-energy CT (P � .67).

Conclusion: Dual-energy CT allows measurement of the degree of en-hancement and detection of calcifications without addi-tional radiation dose.

� RSNA, 2008

1 From the Department of Radiology and Research Insti-tute of Radiology, University of Ulsan College of Medicine,Asan Medical Center, 388-1, Poongnap-dong, Songpa-gu,Seoul 138-736, Korea (E.J.C., J.W.S., J.B.S., Y.M.J.,K.S.S.); and Siemens Medical Solutions, Forchheim, Ger-many (B.K.). Received November 9, 2007; revision re-quested January 14, 2008; revision received April 5; ac-cepted April 17; final version accepted May 29. Sup-ported by a grant (2007-439) from the Asan Institute forLife Sciences, Seoul, Korea. Address correspondence toJ.W.S. (e-mail: [email protected] ).

� RSNA, 2008

ORIGINALRESEARCH

�THORACIC

IMAGING

Radiology: Volume 249: Number 2—November 2008 671

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

The dual-energy (DE) technique of adual-source computed tomographic(CT) scanner allows differentia-

tion of iodine from other materials dueto its stronger photoelectric absorption.CT numbers vary noticeably with beamenergy for materials having a highatomic number, but much less for othermaterials. Therefore, materials can bedifferentiated by applying different x-ray spectra and analyzing the differ-ences in attenuation (1). One potentialapplication of iodine differentiation withthe DE technique is to differentiate con-trast material– enhanced structuresfrom the otherwise dense material inparenchymatous organs, for example,to differentiate complicated cysts fromneoplastic tissue in the kidney (1,2). Todifferentiate calcification from enhanc-ing tissue in a solitary pulmonary nodule(SPN) or in the mediastinal lymph nodeis another potential application.

Chest CT is usually performed as acombination of nonenhanced and en-hanced scanning. A nonenhanced scan isobtained for the detection of calcificationin the SPN or in the lymph node, since thepresence of calcification is one of the im-portant deterministic findings of benig-nity. An enhanced scan is informative inproviding the degree and pattern of en-hancement with use of iodine. In particu-lar, the degree of enhancement of an SPN

after iodine injection has proved to behelpful for distinguishing malignant frombenign nodules (3–9). Application of theDE technique provides a virtual nonen-hanced and an iodine-enhanced imagefrom a single scanning after iodine con-trast material injection by material differ-entiation of iodine. If we are able to detectcalcification on a virtual nonenhanced im-age and directly measure the iodine com-ponent on an iodine-enhanced image, thistechnique may prevent additional nonen-hanced scanning and the patient’s radia-tion exposure will then be reduced. Thepurpose of our study was to assess thepotential clinical utility of DE CT in evalu-ating an SPN in the detectability of calcifi-cation, measurement of the degree of en-hancement, and comparison of the radia-tion dose.

Materials and Methods

A prototype of a commercial softwareproduct Syngo Dual Energy (Siemens Med-ical Solutions) was used for data analysis.B.K. has played a role in data acquisitionand analysis in terms of modification of pa-rameters of dedicated dual-energy softwareand interpretation of acquired CT data.

SubjectsThis research protocol was approved bythe institutional review board of our insti-tution, and written informed consent wasobtained from all patients. From January2007 to July 2007, a total of 49 patients(26 men, 23 women; age range, 31–80years; mean age, 60.39 years � 12.24[standard deviation]) who were sched-uled to undergo percutaneous needle as-piration of an SPN were consecutively en-rolled.

CT ExaminationCT examinations were performed by us-ing the DE mode of the Somatom Defi-nition scanner (Siemens Medical Solu-tions, Forchheim, Germany) with a512 � 512-pixel matrix, 14 � 1.2-mmcollimation, 50 mAs (effective) at 140kV and 210 mAs (effective) at 80 kV,pitch of 0.7, and rotation time of 0.5second. We used a dedicated DE CTprotocol for the thorax, which was rec-ommended by the manufacturer. Tubecurrents of two x-ray tubes of 80 kV and140 kV were fixed as ratio of approxi-mately 4:1 (50 mAs [effective] for 80 kVand 210 mAs [effective] for 140 kV).Each tube has 24 rows of 1.2 mm; how-ever, only some of them were used(14 � 1.2 mm) to allow for an improvedscatter-correction by using the outerrows. The craniocaudal direction wasused for all scans. In patients (n � 25)who had previously undergone a con-ventional chest CT before the proce-dure, the targeted CT scans coveringthe SPN were obtained both before(nonenhanced scan) and 3 minutes aftercontrast material injection (3-minute-delayed scan) for this study (protocol 1)(Fig 1). A 100 mL of iomeprol (Iomeron300; Bracco, Milan, Italy) was adminis-tered at a rate of 3.0 mL/sec by using apower injector. In patients (n � 24)

Published online before print10.1148/radiol.2492071956

Radiology 2008; 249:671–681

Abbreviations:CI � confidence intervalDE � dual energyICC � intraclass correlation coefficientROI � region of interestSPN � solitary pulmonary nodule

Author contributions:Guarantor of integrity of entire study, J.W.S.; study con-cepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manu-script revision for important intellectual content, all au-thors; approval of final version of submitted manuscript,all authors; literature research, E.J.C., J.W.S., J.B.S.,Y.M.J., K.S.S.; clinical studies, E.J.C., J.W.S., J.B.S.,Y.M.J., K.S.S.; statistical analysis, E.J.C., J.W.S., J.B.S.,K.S.S.; and manuscript editing, E.J.C., J.W.S., J.B.S.,B.K., K.S.S.

See Materials and Methods for pertinent disclosures.

Advances in Knowledge

� Dual-energy (DE) CT can provideboth virtual nonenhanced and io-dine-enhanced images from a sin-gle scanning after iodine injection.

� DE CT allows differentiation ofcalcification from enhancing tissueby subtraction of the iodine com-ponent according to the materialdecomposition theory.

� The iodine component in lungnodules is measurable on iodine-enhanced image and is compara-ble to the real value of the degreeof enhancement.

� The radiation dose of a singlescanning covering the full thoraxusing DE CT is not significantlydifferent from that using single-energy CT.

Implications for Patient Care

� DE CT allows measurement of thedegree of enhancement of lungnodules, as well as the detectionof calcifications in the lung nod-ules and lymph nodes.

� By single scanning after iodineinjection with DE CT, we maymeasure the degree of enhance-ment without an additional non-enhanced scan.

THORACIC IMAGING: Solitary Pulmonary Nodules at Dual-Energy CT Chae et al

672 Radiology: Volume 249: Number 2—November 2008

who had not previously undergonechest CT for the evaluation of an SPN, a1-minute-delayed scan covering the fullthorax was acquired for routine clinicalpractice in addition to the protocol 1scan (protocol 2) (Fig 1). For all pa-tients, data of 80 kV, 140 kV, andweighted average image of nonen-hanced and 3-minute-delayed scans tar-geted for SPN were transferred to aworkstation (MultiModality Workplace;Siemens Medical Solutions). Theweighted average image is the approxi-mate 120-kV image, which is automati-cally generated from a combination of140-kV and 80-kV data by using weight-ing factor 1:4 (140 kV: 80 kV). In pro-tocol 2 patients, weighted average im-ages of 1-minute-delayed scan weretransferred to a picture archiving andcommunication system workstation forclinical use.

Material DecompositionBy using two acquisition systems of thedual-source CT, both x-ray tubes can beoperated at different kilovoltage set-tings, allowing the acquisition of DEdata. The DE technique allows differen-tiation and extraction of iodine from theenhanced soft tissue on the basis of thematerial decomposition theory (10,11).Iodine leads to a very large differencebetween the 80-kV and the 140-kV im-age. A diagram shows the CT value at80 kV versus 140 kV, and iodine en-hancement is represented by a vector offixed direction but the length dependson iodine attenuation (Fig 2). As forthese spectra, the typical body materi-als such as fat, water, soft tissue, andliver tissue are approximately locatedon a straight line in this diagram; a two-dimensional linear equation can besolved to obtain the attenuation coeffi-cient (in Hounsfield units) due to iodine(iodine-enhanced image) and an imagethat contains only the contribution ofthe body materials (virtual nonen-hanced image) (12).

Data AnalysisImage data were reconstructed with asection thickness of 2.0 mm by usingkernel D30 and an increment of 2.0mm. By using a modified prototype of

the Liver VNC application class of SyngoDual Energy (Siemens Medical Solu-tions) on a dedicated research MultiMo-dality Workplace workstation, virtualnonenhanced and iodine-enhanced im-ages were calculated from the 3-minute-

delayed scan data. The commerciallyavailable software was modified to allowfor negative CT numbers on the iodineimage, which improves region-of-inter-est (ROI) measurements of very lowconcentrations of contrast material.

Figure 1

Figure 1: Diagram shows the processes of obtaining four series of images at DE CT and comparing theimages and radiation dose. Nonenhanced weighted average and virtual nonenhanced images were comparedfor calcification detection, margin characteristics, image noise, overall diagnostic quality, and CT number ofthe nodules. CT number on iodine-enhanced image and degree of enhancement were also compared. On thediagram, enhanced weighted average image was connected to iodine-enhanced image for simplicity. DSCT �dual-source CT.

Figure 2

Figure 2: Diagram of material decomposition by using the Liver VNC application class of Syngo Dual En-ergy. Virtual nonenhanced and iodine-enhanced images can be generated from reconstruction of single scan-ning at enhanced CT by using the DE technique on the basis of the material decomposition theory.



The material parameters for the mate-rial decomposition method are as fol-lows: �110 HU for fat at 80 kV, �96 HUfor fat at 140 kV, 60 HU for soft tissue at80 kV, and 54 HU for soft tissue at 140kV. It is an empirical observation thatfor DE scans obtained with the Soma-tom Definition scanner, typical tissuesinside the human body are approxi-mately located on this line in the 80-kVversus the 140-kV diagram of theHounsfield units; this includes fat andliver parenchyma, as well as blood,muscle, or water (a similar behavior canbe seen in published results [13] with adifferent scanner and the voltage combi-nation of 100 kV/140 kV). This is alsovalid for mixtures of these materials ona macroscopic or microscopic scale,which seems to apply for lung nodules.

A value of 2.0 was chosen for rela-tive contrast material enhancement,which means that the iodine enhance-ment at 80 kV is twice the value at 140kV. Material decomposition was onlyperformed for voxels with Hounsfieldunits in the mixed image between �300HU (minimum value) and �300 HU(maximum value) (14). The parameter“range,” which controls the size of thespherical three-dimensional filter kernelin units of the voxel size, was set to 2.The maximum value was adjusted to300 from 3071 of the default value be-cause we intended to differentiate be-tween dense calcifications and the io-dine component. The iodine componentwas demonstrated as a color overlay ona background virtual nonenhanced im-age of 3-minute-delayed scan with semi-transparent mode (Fig 3). For each pa-tient, four image series that included anonenhanced weighted average image,enhanced weighted average image, vir-tual nonenhanced image, and iodine-en-hanced image were obtained (Fig 1).

Detectability of Calcification on VirtualNonenhanced ImageOn virtual nonenhanced image, the num-ber and size of calcifications within theSPN and lymph nodes were comparedwith those on nonenhanced weighted av-erage image. A chest radiologist with 5years of experience performed all visualassessments, including detection of calcifi-

cation. Biopsy was performed on all de-tected calcifications within the nodule,other lung nodules within scan rangeand lymph nodes within scan range werecounted. The number of detected calcifi-cations was scored on both virtual nonen-hanced and nonenhanced weighted aver-age images. The calcification size on vir-tual nonenhanced image was comparedwith that on nonenhanced weighted aver-age image according to following grades:grade 1, the same; grade 2, smaller;grade 3, nonvisualized; and grade 4,new calcific attenuation. On enhancedweighted average image, the number ofdetected calcifications was also recordedto determine how many obscured calcifi-cations on enhanced weighted averageimage appear on virtual nonenhanced im-age.

The subjective evaluation of themargin characteristics of an SPN, theimage noise, and the overall diagnosticquality of both virtual nonenhancedand nonenhanced weighted averageimages was performed and comparedwith each other. For margin charac-teristics, grade 1 � smooth; grade 2 �lobulated; grade 3 � irregular; andgrade 4 � spiculated. For the imagenoise of virtual nonenhanced image,grade 1 � lower; grade 2 � the same;and grade 3, greater. For the diagnos-tic quality of virtual nonenhanced im-age, grade 1 � superior; grade 2 � thesame; grade 3 � poor, but diagnostic;and grade 4 � unacceptable.

We measured the chest diameterand recorded the location of the nodulesin all patients to assess the effect of sizeof the small detector system. The loca-tion of the nodules was described ascentral, middle, or peripheral.

Comparison of the CT Number betweenVirtual and Real ImagesA chest radiologist with 5 years of expe-rience placed a circular ROI in as largean area as possible within the SPN. Im-age data of the nonenhanced and the3-minute-delayed scans targeted forSPN were loaded on the “dual energy”tab of the Syngo Dual-Energy applica-tion of the MultiModality Workplaceworkstation. Because two series couldnot be loaded simultaneously, two ROIs

were used to measure the nonenhancedand the 3-minute-delayed scans. To as-sess whether differentiation of iodinewas successful, the CT numbers of SPNon virtual nonenhanced and nonen-hanced weighted average images, aswell as CT numbers of the SPN on io-dine-enhanced image and the degree ofenhancement (CT number on enhancedweighted average image � CT numberon nonenhanced weighted average im-age), were compared. The sum of theCT number of SPN on iodine-enhancedimage and virtual nonenhanced imagewas compared with the CT number onenhanced weighted average image. TheCT numbers of the SPN on iodine-en-hanced image and the degree of en-hancement were compared in terms oftheir diagnostic accuracy for distin-guishing malignant and benign noduleswith a 20-HU threshold. We selected a20 HU as a cutoff value on the basis ofresults of previous studies (3–6).

Radiation DoseThe radiation doses of protocol 1 (bothnonenhanced and 3-minute-delayed scanstargeted for SPN, n � 25) and protocol2 (protocol 1 plus 1-minute-delayedscan covering full thorax, n � 24) in allpatients were recorded as the value ofthe dose-length product. Dose parame-ters were obtained from the “PatientProtocol” image (Siemens Medical Solu-tions), which is an image file that in-cludes the information of the CT proto-col and the amount of radiation dose.That image is automatically exported topicture archiving and communicationsystem workstation with other CT im-ages. Dose-length product of the1-minute-delayed scan covering the fullthorax in 24 patients was used for com-parison with other examinations usingsingle-energy CT. The radiation dosewas recorded in 24 CT examinationswith 16-detector CT (Somatom Sensa-tion 16, Siemens Medical Solutions) andin 24 examinations with dual-source CTwithout applying the DE mode. Withoutthe DE mode, only the A system oper-ates; therefore, this works as same asthe 64-detector CT. Those 48 CT exam-inations were randomly selected amongall the CT examinations performed on 3



consecutive days in our institution. Theradiation doses of multidetector CT,dual-source CT, and DE CT were thencompared.

Statistical AnalysisCommercial statistical packages (ver-sion 12.1.1 SPSS, Chicago, Ill; and ver-sion 7.3, MedCalc Software, Mari-akerke, Belgium) were used. The re-sults were expressed as mean �

standard deviation. Descriptive statis-tics were used to evaluate the numberand size of the calcifications, imagenoise, and overall diagnostic quality.Agreement of the margin characteris-tics of SPN between virtual nonenhancedimage and nonenhanced weighted averageimage was analyzed with � statistics.Agreement in the CT numbers of SPNon four image series was analyzed bycalculating the intraclass correlation co-

efficient (ICC) and using the Bland-Altman method. The Mann-Whitney Utest was used to analyze statistically sig-nificant differences between the CTnumbers of benign and malignant nod-ules. Between the CT number on iodine-enhanced image and the degree of en-hancement, the diagnostic accuracy wascompared by calculating the sensitivity,specificity, and accuracy with 95% con-fidence intervals (CIs) (15). The Mann-

Figure 3

Figure 3: Images in 57-year-old man with a nodule. (a) Nonenhanced weighted average image. (b) Virtual nonenhanced image. (c) Iodine-enhanced image. The io-dine component was demonstrated as a color overlay in the semitransparent mode. (d) Enhanced weighted average image. (e) Nonenhanced weighted average imageshows an ROI in the nodule with a CT number of 21.8 HU. (f) VNC of ROI represents the CT number on background virtual nonenhanced image and OL represents the CTnumber on iodine-enhanced image, which indicates the iodine component in the nodule. The CT number (20.9 HU) of VNC approximates the CT number (21.8 HU) of thenodule on nonenhanced weighted average image. A circle represents the field of view of the smaller detector system.



Whitney U-test was used to comparethe radiation dose among the CT exam-inations. A P value of less than .05 wasconsidered to indicate a significant dif-ference, and all P values were twotailed.

Results

The size of SPNs ranged from 5 to 70mm, and the average size was 24.8 mmin diameter � 11.8. The average ROIwas 705.7 pixels � 962.2 and 1.58cm2 � 2.3. The average maximum chestdiameter of our patients was 304.56

mm � 25.11. Among the central, mid-dle, and peripheral nodule locations, 11nodules were located within the periph-ery.

Detectability of Calcification on VirtualNonenhanced ImageOn nonenhanced weighted averageimage, 20 calcifications were presentin the nodules of four patients and 45calcifications were present in the lymphnodes of 11 patients (Figs 4 and 5). Onvirtual nonenhanced image, 85.0% (17of 20) of calcifications in the nodulesand 97.8% (44 of 45) of calcifications in

the lymph nodes were detected. On vir-tual nonenhanced image, the sizes of thecalcifications were smaller than thoseon nonenhanced weighted average im-age in all patients with calcification(n � 14).

The margin characteristics betweenvirtual nonenhanced image and nonen-hanced weighted average image showeda strong agreement (weighted � in thequadratic set, 0.91; standard error,0.14). Virtual nonenhanced image hadthe same degree of noise in 16.3%(eight of 49) of our patients and morenoise in 83.7% (41 of 49) of these pa-

Figure 4

Figure 4: Images in 52-year-old woman with adenocarcinoma. (a) Enhancedweighted average image shows multiple calcifications with variable high attenua-tion in the mass in the right upper lobe. (b) Nonenhanced weighted average image.(c) Virtual nonenhanced image.



tients. Virtual nonenhanced image hadthe same diagnostic quality in 77.6%(38 of 49) of patients, inferior but diag-nostic quality in 16.3% (eight of 49),and unacceptable quality in 8.2% (fourof 49). In eight patients, inferior butacceptable diagnostic image quality wascaused by decreased grading of the mar-gin characteristics. The unacceptablediagnostic image quality in four patientsoccurred because the number of de-tected calcifications on virtual nonen-hanced image was lower than that onnonenhanced weighted average image.In two patients, four calcifications thatwere obscured on enhanced weightedaverage image were depicted on virtualnonenhanced image (Fig 6).

Comparison of the CT Number on Virtualand Real ImagesCT numbers of SPN on nonenhancedweighted average and virtual nonen-hanced images showed good agreement(ICC, 0.83; P � .001). The CT numberof SPN on iodine-enhanced image andthe degree of enhancement (CT numberon enhanced weighted average imageminus CT number on nonenhancedweighted average image) also showedgood agreement (ICC, 0.91; P � .001).The sum of the CT numbers of SPN on

iodine-enhanced image and virtual non-enhanced image showed a strong agree-ment with that on enhanced weightedaverage image (ICC, 0.98; P � .001).Results of the Bland-Altman method ofthe agreement of CT numbers of SPN onnonenhanced weighted average and vir-tual nonenhanced images, as well as theCT number of SPN on iodine-enhancedimage and the degree of enhancement,are shown in Figure 7. The mean differ-ence in CT numbers of SPN on nonen-hanced weighted average image and vir-tual nonenhanced image was 4.3 HU(95% CI: �14.6 HU, 23.3 HU), while itwas �0.5 HU (95% CI: �17.6 HU, 16.5HU) between CT number of SPN on io-dine-enhanced image and the degree ofenhancement. The mean difference inCT numbers on enhanced weighted av-erage image and the sum of CT numberson iodine-enhanced image and virtualnonenhanced image was 3.8 HU (95%CI: �5.3 HU, 12.9 HU).

A total of 45 of 49 nodules wereconfirmed as benign or malignant onthe basis of percutaneous needle aspi-ration findings, and four nodules werenot confirmatively diagnosed. Theprevalence of malignancy was 55.6%(25 of 45 nodules), and diagnoses in-cluded adenocarcinoma (n � 19),

squamous cell carcinoma (n � 2), me-tastasis (n � 2), lymphoma (n � 1), andlarge cell neuroendocrine cell carcinoma(n � 1). Diagnoses of benign nodulesincluded tuberculosis (n � 7), hamar-toma (n � 3), sclerosing hemangioma(n � 2), abscess (n � 1), aspergilloma(n � 1), calcified granuloma (n � 1),and a nonspecific inflammatory lesion(n � 5). Any nondiagnostic result suchas insufficient specimen was notpresent.

The results of both the CT numberon iodine-enhanced image and the de-gree of enhancement showed that ma-lignant nodules enhance significantlymore than do benign nodules (P � .001for iodine image; P � .001 for degree ofenhancement) (Table 1). By comparingthe diagnostic accuracy with a cutoffvalue of 20 HU for malignant nodules,the CT number of SPN on iodine-en-hanced image had higher sensitivity andaccuracy than did the number using thedegree of enhancement (Table 2).

Radiation DoseCT examinations using protocol 1 wereperformed in 25 patients and those us-ing protocol 2 were performed in 24patients. The average dose-length prod-uct was 92.77 mGy � cm � 32.37 in pro-

Figure 5

Figure 5: Depiction of calcification in the mediastinal lymph nodes. (a) Nonenhanced weighted average image shows two calcifications in the right lower paratrachealnodal station. (b) Virtual nonenhanced image shows two calcifications in the lymph node. The size of calcifications was smaller on this image. (c) Enhanced weightedaverage image.



tocol 1 and 333.54 mGy � cm � 53.93 inprotocol 2. The average dose-lengthproduct of 1-minute-delayed scan cov-ering the full thorax in protocol 2 was240.77 mGy � cm � 37.18. This was notsignificantly different from the averagedose-length product of CT examinationsperformed by using single-energy multi-detector CT and dual-source CT without

the application of the DE mode(235.38 mGy � cm � 35.40 and 233.13mGy � cm � 67.36, respectively) (P �.67 and .39, respectively).

Discussion

On the basis of our study findings, weascertained that the iodine component

could be successfully differentiatedfrom the enhanced soft tissue at en-hanced CT using DE. CT numbers onnonenhanced weighted average imageand virtual nonenhanced image, as wellas the CT number on iodine-enhancedimage and the degree of enhancement,showed reliable agreement. The evalu-ation of tumor vascularity by measuring

Figure 6

Figure 6: Example of an obscured calcification on enhanced weighted averageimage that was detectable on virtual nonenhanced image. (a) Nonenhancedweighted average image shows an ill-defined calcification in right paratracheallymph node. Another dense calcified plaque is seen in the posterior wall of theaorta. (b) Virtual nonenhanced image well depicts two calcifications in the lymphnode and aorta. The calcification in the lymph node became blurred probably due tonoise. (c) On enhanced weighted average image, the calcification in the lymph nodeis not visible due to enhancement.



the degree of enhancement of SPN afteriodine contrast agent administrationhas proved to be helpful for distinguish-ing malignant nodules from benign nod-ules at contrast-enhanced dynamic CT(3–9). However, contrast-enhanced dy-namic CT has the disadvantage of anincreased radiation dose to patients andof inaccurate measurement due to thedifferent positioning of the ROI in theSPN on the serial images of dynamicscanning. A DE CT simultaneously pro-vides virtual nonenhanced image and aniodine-enhanced image from a singlescanning; therefore, we can acquire theCT number on a nonenhanced imageand the iodine value in the SPN from

the same ROI. This technique may pre-vent a nonenhanced scan as baselinestudy for dynamic enhancement studyof nodules and it may also reduce mea-

surement error due to different posi-tioning of the ROI on sequential images.Effect on measurement error might bemore important in the relatively smaller

Figure 7

Figure 7: Graphs depict results of the Bland-Altman method. (a) Nonenhancedweighted average image and virtual nonenhanced image (mean difference, 4.3 HU;95% CI: �14.6 HU, 23.3 HU). (b) The degree of enhancement and CT number oniodine-enhanced image (mean difference, �0.5 HU; 95% CI: �17.6 HU, 16.5 HU).(c) Enhanced weighted average image and virtual nonenhanced image plus iodine-enhanced image (mean difference, 3.8 HU; 95% CI: �5.6 HU, 12.9 HU). SD � stan-dard deviation.

Table 1

Comparison of the Measurement of Enhancement

Analyzed DataMalignant Nodules(n � 25)

Benign Nodules(n � 20) P Value

CT number on iodine-enhanced image (HU) 36.5 � 16.0 17.3 � 21.8 .001Degree of enhancement (HU) 37.0 � 14.6 17.0 � 17.9 �.001

Note.—Data are mean � standard deviation.



nodules. A single, enhanced CT scan-ning using DE provides additional infor-mation about the degree of enhance-ment of an SPN without additional radi-ation exposure.

In our study, there was a tendencyfor the sum of the CT numbers on vir-tual nonenhanced image and on iodine-enhanced image to be smaller than theCT number on enhanced weighted aver-age image. Theoretically, the sum of CTnumbers of those two calculated imageshas to be the same as the CT number onenhanced weighted average image whenthe DE technique is used to decomposeenhanced structures into soft tissue andiodine. However, this is only valid aslong as virtual nonenhanced, iodine-en-hanced, and enhanced weighted aver-age images are derived for the samevolume; in reality, virtual nonenhancedimage and iodine-enhanced image in-clude a slightly larger volume due tofiltering before material decomposition.Therefore, this effect may be caused bypartial volume effects with air that werelarger after application of the filtering.

Results of the Bland-Altman methodand the ICC showed that the degree ofenhancement and the iodine-enhancedimage have a stronger agreement thando nonenhanced weighted average andvirtual nonenhanced images. Moreover,the standard deviation of the degree ofenhancement minus iodine-enhancedimage is smaller than the standard devi-ation of nonenhanced weighted averageimage minus virtual nonenhanced im-age. This should not happen if an en-hanced weighted average image isequivalent to virtual nonenhanced im-age plus iodine-enhanced image, butcan again be explained by partial vol-ume effects with air. In general, partialvolume effects with air will have a big-ger effect on virtual nonenhanced image

than on iodine-enhanced image, as theresulting shift of the CT numbers in the80-kV versus 140-kV diagram is almostparallel to the difference vector be-tween soft tissue and fat. The lower in-fluence of partial volume effects on theiodine-enhanced image is beneficial forpractice, because the CT number on theiodine-enhanced image and not on thevirtual nonenhanced image is of inter-est, that is, is used for evaluating tumorvascularity. In principle, the CT num-bers on virtual nonenhanced and iodine-enhanced images are expected to havelarger systematic errors than the CTnumbers on nonenhanced weighted av-erage and enhanced weighted averageimages, because the material decompo-sition is sensitive to systematic errorson the 80-kV image, which—in compar-ison with the 140-kV image—is moreaffected by beam hardening artifactsand artifacts due to cross-scattered ra-diation. However, in reality, the erroron iodine-enhanced image seems to beof similar magnitude or smaller than theerror on the measurement of enhancedweighted average image minus nonen-hanced weighted average image, whichresults from enhanced weighted aver-age image and nonenhanced weightedaverage image being measured at twodifferent ROIs. We evaluated the agree-ment by two ways: the Bland-Altmanmethod and the ICC. Although therewas some mismatch between the resultsof the two methods, the overall agree-ment was good. Moreover, in terms ofdiagnostic accuracy for malignancy, theCT number on the iodine-enhanced im-age showed a comparable result withthe real degree of enhancement.

Demonstration of calcification onthe virtual nonenhanced image, how-ever, was not comparable to that on thenonenhanced weighted average image,

since most calcifications appearedsmaller on virtual nonenhanced imagethan on nonenhanced weighted averageimage. Although most of the calcifica-tions (93.8%) in the nodules and lymphnodes were detected on virtual nonen-hanced image, some were not (6.2%).There are several reasons for this re-sult: image noise of the virtual nonen-hanced image is higher; the signal-to-noise ratio in the surface of calcifica-tions is lower; the low-pass filter blursthe edges; and during material decom-position, the signal is lowered becausepart of the calcium is moved to the io-dine image. Since presence of calcifica-tion in the lung nodules and lymphnodes implies benign etiology, differen-tiating calcification from contrast en-hancement is important. However,some institutions, including our hospi-tal, perform only enhanced scanningbecause of the relatively low benefit ofa nonenhanced scanning, especiallyconsidering its double radiation dose.Enhancement of an SPN after iodineinjection often obscures the presenceof calcification in the nodule, espe-cially if the calcification does not showsubstantially higher attenuation thanthe peak level of enhancement of thenodule. In this study protocol, we ob-tained 3-minute-delayed scans, asmost malignant lung nodules show thepeak level of enhancement at approx-imately 3 minutes (3,9). In our study,obscured calcifications on enhancedweighted average image were de-tected on virtual nonenhanced imagein two patients. This result impliesthat a virtual nonenhanced imagecould be helpful when high attenuationis equivocal on an enhanced CT image.When high attenuation is suspected,we can retrospectively acquire a vir-tual nonenhanced image by means ofpostprocessing after CT by using DE.

The radiation dose of 1-minute-de-layed scans covering the full thorax byusing a DE CT was similar to that ofscanning by using single-energy multide-tector CT or dual-source CT without theDE mode. This result indicates that sin-gle scanning by using a DE mode of adual-source CT saves the amount of ra-diation of an additional nonenhanced

Table 2

Comparison of the Diagnostic Accuracy for Malignancy with a Cutoff Value of 20 HU

Analyzed Data Sensitivity (%) Specificity (%) Accuracy (%)

CT number on iodine-enhanced image 92.0 (52.2, 85.9) 70.0 (47.9, 85.7) 82.2 (56.5, 82.4)Degree of enhancement 72.0 (73.9, 98.9) 70.0 (47.9, 85.7) 71.1 (68.4, 90.7)

Note.—Data in parentheses are 95% CIs.



scanning. In routine practice, optionalreconstruction of a virtual nonenhancedimage after DE CT scanning may re-place the additional nonenhanced CTscan.

One of the limitations of dual-sourceCT was the 26-cm field of view of thesmaller detector system, unlike the50-cm field of view of the larger detec-tor system (Fig 3e and 3f). The regioninside the field of view of the smallerdetector system can be decomposed.Because of the 26-cm field of view of thesmaller detector system, the peripheralportion of the lower lung could not becovered in some of our subjects. There-fore, patients with a small nodule in thefar periphery of the lower lungs cannotbe scanned by using this protocol. In thepresent study, all lung nodules were lo-cated inside the field of view of thesmaller detector.

Another limitation of our study wasthat only a small number of patientshaving an SPN with calcification wereincluded, since most of our study pa-tients were planning to undergo percu-taneous needle aspiration due to thepresence of a radiologically indetermi-nate nodule. Detected calcifications inour study were mostly in the lymphnodes. One of the limitations was thesmall number of patients, which demon-strates that further studies are neededto definitely establish the feasibility ofthe new tool. Therefore, this applicationawaits further investigation to deter-

mine its reliability for detecting calcifi-cation in an SPN.

In conclusion, we were able to suc-cessfully differentiate the iodine compo-nent from the enhanced soft tissue byusing DE CT. Acquisition of virtual non-enhanced and iodine-enhanced imageswith DE CT may be a feasible tool forthe measurement of the degree of en-hancement.

Acknowledgments: We acknowledge Park Joo-Young for her technical assistance and also thankBonnie Hami, MA, for her contribution and adviceregarding the manuscript editing.

References1. Johnson TR, Krauss B, Sedlmair M, et al.

Material differentiation by dual energy CT:initial experience. Eur Radiol 2007;17:1510–1517.

2. Flohr TG, McCollough CH, Bruder H, et al.First performance evaluation of a dual-source CT (DSCT) system. Eur Radiol 2006;16:256–268.

3. Swensen SJ, Morin RL, Schueler BA, et al.Solitary pulmonary nodule: CT evaluation ofenhancement with iodinated contrast mate-rial—a preliminary report. Radiology 1992;182:343–347.

4. Swensen SJ, Brown LR, Colby TV, WeaverAL. Pulmonary nodules: CT evaluation of en-hancement with iodinated contrast material.Radiology 1995;194:393–398.

5. Yamashita K, Matsunobe S, Tsuda T, et al.Solitary pulmonary nodule: preliminarystudy of evaluation with incremental dy-namic CT. Radiology 1995;194:399–405.

6. Swensen SJ, Brown LR, Colby TV, Weaver

AL, Midthun DE. Lung nodule enhancementat CT: prospective findings. Radiology 1996;201:447–455.

7. Zhang M, Kono M. Solitary pulmonarynodules: evaluation of blood flow patternswith dynamic CT. Radiology 1997;205:471–478.

8. Swensen SJ, Viggiano RW, Midthun DE,et al. Lung nodule enhancement at CT: mul-ticenter study. Radiology 2000;214:73–80.

9. Jeong YJ, Lee KS, Jeong SY, et al. Solitarypulmonary nodule: characterization withcombined wash-in and washout features atdynamic multi-detector row CT. Radiology2005;237:675–683.

10. Brooks RA. A quantitative theory of theHounsfield unit and its application to dualenergy scanning. J Comput Assist Tomogr1977;1:487–493.

11. Kelcz F, Joseph PM, Hilal SK. Noise consid-erations in dual energy CT scanning. MedPhys 1979;6:418–425.

12. Kalender WA, Klotz E, Kostaridou L. Analgorithm for noise suppression in dual en-ergy CT material density images. IEEE TransMed Imaging 1988;7:218–224.

13. Zatz LM. The effect of the kVp level on EMIvalues: selective imaging of various materialswith different kVp settings. Radiology 1976;119:683–688.

14. Siegel MJ, Schmidt B, Bradley D, Suess C,Hildebolt C. Radiation dose and image qual-ity in pediatric CT: effect of technical factorsand phantom size and shape. Radiology2004;233:515–522.

15. Medina LS, Zurakowski D. Measurementvariability and confidence intervals inmedicine: why should radiologists care? Ra-diology 2003;226:297–301.



clinical utility of dual-energy ct in the evaluation of ... 3 - new... · gle-energy ct (p.67)....

Documents