Carotid bifurcation calcium and correlation with percent stenosis of the internal carotid artery on CT angiography

Alexander M. McKinney

Sean O. Casey

Mehmet Teksam

Leandro T. Lucato

Maurice Smith

Charles L. Truwit

Stephen Kieffer

Carotid bifurcation calcium and correlationwith percent stenosis of the internal carotidartery on CT angiography

Received: 30 December 2003Accepted: 4 September 2004Published online: 14 January 2005� Springer-Verlag 2005

Abstract The aim of this paper wasto determine the correlation betweencalcium burden (expressed as a vol-ume) and extent of stenosis of theorigin of the internal carotid artery(ICA) by CT angiography (CTA).Previous studies have shown thatcalcification in the coronary arteriescorrelates with significant vessel ste-nosis, and severe calcification (mea-sured by CT) in the carotid siphoncorrelates with significant (greaterthan 50% stenosis) as determinedangiographically. Sixty-one patients(age range 50–85 years) underwentCT of the neck with intravenousadministration of iodinated contrastfor a variety of conditions. Imageswere obtained with a helical multi-detector array CT scanner andreviewed on a three-dimensionalworkstation. A single observermanipulated window and level tosegment calcified plaque fromvascular enhancement in order toquantify vascular calcium volume(cc) in the region of the bifurcation ofthe common carotid artery/ICA ori-gin, and to measure the extent ofICA stenosis near the origin. A totalof 117 common carotid artery bifur-cations were reviewed. A‘‘significant’’ stenosis was definedarbitrarily as >40% (to detectlesions before they become hemody-namically significant) of luminaldiameter on CTA using NASCET-like criteria. All ‘‘significant’’stenoses (21 out of 117 carotidbifurcations) had measurable

calcium.We found a relatively strongcorrelation between percent stenosisand the calcium volume (Pearson’s r= 0.65, P<0.0001). We also foundthat there was an even stronger cor-relation between the square root ofthe calcium volume and the percentstenosis as measured by CTA (r=0.77, P<0.0001). Calcium volumesof 0.01, 0.03, 0.06, 0.09 and 0.12 ccwere used as thresholds to evaluatefor a ‘‘significant’’ stenosis. A re-ceiver operating characteristic(ROC) curve demonstrated thatthresholds of 0.06 cc (sensitivity88%, specificity 87%) and 0.03 cc(sensitivity 94%, specificity 76%)generated the best combinations ofsensitivity and specificity. Hence, thispreliminary study demonstrates arelatively strong relationship be-tween volume of calcium at the car-otid bifurcation in the neck(measured by CT) and percent ste-nosis of the ICA below the skull base(as measured by CTA). Use of cal-cium volume measurements as athreshold may be both sensitive andspecific for the detection of signifi-cant ICA stenosis. The significanceof the correlation between calciumvolume and ICA stenosis is thatpotentially a ‘‘score’’ can be obtainedthat will identify those at risk forhigh grade carotid stenosis.

Keywords Calcium Æ Carotidarteries Æ Calcification Æ Computedtomography (CT) Æ CT angiography(CTA) Æ Carotid stenosis

Neuroradiology (2005) 47: 1–9DOI 10.1007/s00234-004-1301-4 DIAGNOSTIC NEURORADIOLOGY

Presented at the 41st Annual Meeting of theAmerican Society of Neuroradiology,Washington D.C., 2003. Sean Casey, MDand Charles Truwit, MD are members ofthe Medical Advisory Board of Vital Ima-ges (Plymouth, Minnesota), the companythat developed the Vitrea 2 software.

A. M. McKinney (&) Æ S. O. CaseyM. Teksam Æ C. L. Truwit Æ S. KiefferDepartments of Radiology, University ofMinnesota Medical School, and HennepinCounty Medical Center, Minneapolis,MN 55455, USAE-mail: [email protected].: +1-612-2734092Fax: +1-612-6248495

L. T. LucatoClinics Hospital of the University of SaoPaulo, Sao Paulo, Brazil

M. SmithDepartment of Biomedical Engineering,Johns Hopkins University, Baltimore,MD 21218, USA

Introduction

Atherosclerotic disease in the coronary and carotidarteries is a significant cause of morbidity and mortality[1]. Coronary and carotid artery atherosclerotic diseaseshare similar risk factors and are often coexistent in theat-risk population, especially diabetics, hypertensivepatients, smokers, patients with hypercholesterolemiaand those of advanced age [2–4]. Calcification in thecoronary arteries has been correlated with significantvessel stenosis (defined as greater than 50%) [5]. Severecalcification in the carotid siphon has also been associ-ated with a greater than 50% carotid stenosis [6]. Cor-onary calcium scoring has been implemented for earlydetection of stenoses in high risk or symptomatic pa-tients, and usually focuses on thresholds of densitymeasurements obtained by helical CT or electron beamCT (EBCT) [5, 7, 8], although this technique hasencountered controversy due to being sensitive (88%)but not quite as specific (52%) [5].

Ultrasound and MRA are commonly used non-inva-sive methods to evaluate for carotid stenosis in an at-riskpopulation, and both modalities have been demonstratedto be sensitive for the detection of carotid stenosis, al-though ultrasound has been described as slightly lessaccurate and specific than MRA, particularly in patientswith less than 70% stenosis [9, 10]. CTA has a provenstrong correlation with percent stenosis in evaluating forinternal carotid artery (ICA) stenosis near the carotidbifurcation, relative to the gold-standard conventionalcatheter angiography [11–13]. However, CTA is nottypically used as the initial evaluation for carotid stenosisdue to expense and risks associated with intravenousadministration of iodinated contrast material.

The primary goal of this study was to test thehypothesis that a sample volume of calcium at the car-otid bifurcation may correlate with percent luminal ste-nosis. CTA was used as the relative gold standard in ourstudy to evaluate the percent of luminal internal carotidstenosis near the bifurcation. As a preliminary study todetermine the feasibility of using carotid calcium volumeto detect ‘‘significant’’ stenoses, we studied a series ofroutine post-intravenous contrast neck CT examinationsperformed on a multidetector CT scanner. The data setsobtained from these scans allowed for simultaneouscalcium volume measurements while also including theCT angiographic data for stenosis measurements.

Materials and methods

Sixty-one consecutive patients (age range 50–85 years,mean 63 years) were included in the study, all of whomunderwent clinically indicated neck CT with a helicalmultidetector array (MDA) CT scanner following abolus of intravenous administration of iodinated con-

trast. Age (equal to or greater than 50 years) was theonly significant factor for inclusion in this study. Pa-tients were not subselected for diabetes or hypertensionor other risk factors known to accelerate atheroscleroticdisease. These examinations were performed for variousindications, mostly for the purpose of evaluating headand neck cancer. None of the examinations was per-formed for suspected carotid occlusive disease. Thirty-eight females and 23 males were included. One patientwith a known large invasive neck malignancy was ex-cluded on the side of carotid encasement (right side),since it was difficult to distinguish carotid lumenthroughout most of the neck. In four patients withknown carotid endarterectomy, the side of involvementwas also excluded on the side of surgery (two on theright, two on the left), since we considered the extent ofdisease to be unrepresentative of the true or past calciumburden; however, we included the non-operative side inall four cases. Hence, a total of five carotid bifurcationswere excluded (out of 122 possible carotids in 61 pa-tients).

All patients were evaluated with post-contrast neckCT studies, which were obtained on a quad detectorarray system (Siemens Volume Zoom; Forcheim, Ger-many). Scanning was performed using 120 kVp with amaximum tube current of 230 mA (adjustable effectivemA depending on patient body habitus). Scanning timewas approximately 15 s, and images were obtainedduring a 15–20 s breath hold. Scans were performed in acaudal-cranial direction from the level of the aortic archup to 1 cm above the level of the sella in order to includethe Circle of Willis. Source images were displayed with a20–24 cm field of view depending on patient body hab-itus. A 22 gauge catheter or larger was used for injectionof contrast medium in peripheral veins in either theantecubital region, forearm, or hand, for a total of150 cc. Pre-intravenous contrast scans were not ob-tained. Non-ionic contrast material (ioversol 320 mg/ml;Mallinckrodt, St Louis, Miss., USA) was administeredin the following fashion: 50 cc were injected manuallyprior to obtaining a scout topogram, and then scanrange was set. Within minutes after the manual injec-tion, the power injector was connected to the catheter,and contrast infusion started at a rate of 3 cc/s for thefirst 70 cc and 2 cc/s for the last 30 cc. Delay betweeninjection and scan was 30 s. If significant dental amal-gam was noted on the scout topogram, power injectionrate was 3 cc/s for the first 40 cc, 1.5 cc/s for the next35 cc, and after a 20 s delay (and reangling of scan planethrough the region of the mouth), the last 25 cc waspower injected at a rate of 2.8 cc/s, 1 mm collimationwas used to produce 4 mm thick images for routineclinical use. Raw data were then retrospectively recon-structed to 1 mm thick sections at 0.5 mm intervals foruse in calcium volume measurement and CTA evalua-tion of the carotid arteries.

2

Fig. 1 a and b Maximumintensity display CTA coronalview (a) and sagittal view of theleft side (b) demonstrates thecarotid bifurcation calciumbilaterally (yellow arrows).There is also a small amount ofcalcified plaque more cephaladin the right internal carotidartery (red arrows). Decentquality CT angiographic imagessuch as these were obtainablefrom the multidetector neck CTraw data. Note that the carotidcalcifications are easily distin-guishable from the arterial lu-men due to the higherHounsfield unit range of cal-cium. c (right and left carotids)Prior to carotid calcium mea-surement, the 3D data set wassculpted to a subvolume toexclude bony calcium related tospine, thyroid, and hyoid carti-lage. d A density threshold wasthen visually adjusted (within arange of window 10–20 HUand level 250–500 HU) to in-clude only calcium within thevolume (thus excluding luminalcontrast enhancement). Volumemeasurements were then calcu-lated with an automated featureof the software. This patient hasa left carotid calcium volume of0.13 cc, and a right calciumvolume of 0.10 cc. Note thesmall amount of calcium morecepahalad in the right ICA (redarrow). e Measurements of ste-notic right and left ICA luminaldiameter. CTA stenosis mea-surements were obtained in aNASCET-like fashion by mea-surement of ICA luminaldiameter at the point of maxi-mum stenosis (same patient asa–d). Non-calcified plaque (yel-low arrow) was distinguishablefrom calcified plaque in thispatient. f Measurements of thenormal right and left ICAluminal diameter. At a level2 cm or greater cephalad to themaximal stenosis, the morenormal ICA diameter is ob-tained. This patient had anapproximately 45% stenosis onthe right, and 43% on the left

3

The reconstructed raw data were then sent to a Vitrea2 workstation (Vital Images, Plymouth, Minn., USA)for evaluation of CTA and quantification of total vas-cular calcium volume by a single observer. The observerwas able to view vessels in an infinite number of multi-planar views, including oblique angulations. The ob-server would first evaluate using various windows andlevels for mass, trauma or other extravascular etiologythat could alter visualization or measurement of thecarotid bifurcation (Fig. 1a, b).

For the purpose of carotid bifurcation calcium bur-den measurement, a volume was manually ‘‘sculpted’’from 2 above to 2 cm below the carotid bifurcationwhile taking care not to include thyroid cartilage, cricoidcartilage, or styloid process calcifications (Fig. 1c). This‘‘sculpting’’ consisted of drawing the area intended forinclusion on serial axial images, so that other sources ofcalcification/bone (e.g. thyroid/cricoid cartilage andvertebra) were outside of the drawn area and hence ex-cluded from the resulting sculpted volume. Externalcarotid calcification near the bifurcation was not sepa-rated from that in the internal carotid artery whilesculpting. After sculpting, the window and level werethen manually altered to maximize the amount of visu-alized calcium (level 250–500 HU, window 10–20 HU)while not including vascular contrast (Fig. 1d). Thiscombination of sculpting and windowing would result invisualization of only the bright/dense bifurcation cal-cium. Using a volume calculation function available onthe workstation, plaque burden was calculated in cc. Thesoftware would automatically calculate the remainingvolume (in cc) of the visualized calcium by the press of abutton. The smallest plaque volume that can be calcu-lated by the software is 0.01 cc.

After excluding extraneous cause of carotid involve-ment (such as tumor encasement/invasion), bifurcationvascular anatomy was evaluated by manipulating win-dow/level (window 900–1200 HU, level 300–500 HU) toview calcified and non-calcified plaque and measureinternal carotid stenoses. Calcified plaque was easilydistinguished from non-calcified plaque by manuallyand visually manipulating window/level. The singlereader was not blinded to results of calcium volumedetermination before measuring degree of stenosis. Allmeasurements of stenosis were obtained on 1 mmthickness images to minimize volume averaging. Toavoid overestimation of luminal diameter, the imagesused to measure luminal diameter stenosis were alwaysobtained from an image plane perpendicular (90�) to theaxis of the vessel. This often required reorientation ofthe image plane. Vascular luminal contrast enhancementwas easily separated visually from calcified plaque (ifpresent) by manual manipulation of window/level set-tings. This manipulation was monitored closely to avoidaccidental inclusion of vascular enhancement in thecalcium volume. To evaluate for the presence of stenosis,

the observer measured the cross-sectional luminaldiameter in millimeters at the narrowest location(Fig. 1e) (even in the absence of significant plaque) andmeasured the more normal/non-stenotic internal carotidartery diameter at least 2 cm above the site of narrow-ing/plaque (Fig. 1f). Thus, an attempt was made to ad-here to North American Symptomatic CarotidEndarterectomy Trial (NASCET) criteria as closely aspossible [14] when performing stenosis measurements.Locations of narrowing in the ICA were evaluated atany level from the carotid bifurcation to the skull base;common carotid artery narrowing and intracranial ICAnarrowing were not evaluated in this study. Images ofthe final calcium burden volume and measurements ofluminal diameter were stored for future reference. Asingle observer initially evaluated all studies, although asecond observer re-evaluated those patients with ‘‘sig-nificant’’ stenoses.

For statistical analysis, the sensitivity and specificityas well as a receiver operating characteristic (ROC)curve of the study data were obtained at five arbitraryvolume thresholds of calcium burden: 0.01, 0.03, 0.06,0.09, and 0.12 cc. We arbitrarily used 40% as ourdefinition for ‘‘significant stenosis’’ in order to detect astenotic lesion before the stenosis reached 50% (whichis commonly considered the threshold at which a ste-nosis is hemodynamically significant). We considered apositive test one in which the patient had a greaterthan 40% luminal stenosis in one or both of the ICAsin the presence of the calcium burden threshold atwhich we were testing. A Pearson’s correlation coeffi-cient (r) was calculated to assess for degree of corre-lation between calcium volume and percent stenosis.The correlation coefficient (r) can range from )1.0 to1.0, where )1.0 is a perfect negative (inverse) correla-tion, 0.0 is no correlation, and 1.0 is a perfect positivecorrelation. Typically, an r of 0.4–0.7 is considered amoderate correlation and 0.7–0.9 is a strong correla-tion. The statistical significance of r was tested using at-test. A statistician reviewed the data and these cal-culations.

Results

Vascular calcification was not difficult to distinguishand segment from luminal carotid contrast enhance-ment on the workstation. Total time to measure cal-cium burden and stenosis was approximately 2–5 minper side, depending largely on the tortuosity of thevessel and the degree of calcification. Thirty-five of the61 patients in this study demonstrated measurablecarotid bifurcation calcium burden greater than 0 cc(27 on the right, 30 on the left), 21 of whom hadbilateral calcification. We noted that the computertruncated two cases of very tiny amounts of visualized

4

calcium (apparent to the observer) to 0 cc. Carotidbifurcation calcium ranged from 0 to 0.76 cc. Themean calcium burden of all carotid bifurcations was0.053 cc (SD 0.112).

CTA measures of ICA stenoses ranged from 0 to99% (mean 14.3%). All measurable stenoses (38 car-otids in 25 patients with a stenosis >0%) except onehad detectable calcium, and this exception had a ste-nosis of only 19%. Nevertheless, this patient hadmeasurable calcium on the opposite side (0.13 cc with38% stenosis on the opposite side). The greatest car-otid bifurcation calcium volume without measurablestenosis by NASCET criteria was 0.17 cc. A ‘‘signifi-cant’’ stenosis (which we defined as > 40%) was seenin 16 patients (for a total of 21 carotids), where fivepatients had bilateral ‘‘significant’’ stenoses. For the‘‘significant’’ stenoses, the mean calcium burden was0.215 cc (SD 0.204). All ‘‘significant’’ stenoses dem-onstrated a measurable calcium burden, although onepatient with prior carotid endarterectomy had a 99%stenosis and only mild calcium burden (0.04 cc). Thiswas one of two cases of 99% stenosis in this study,both of which had previously undergone prior carotidendarterectomy on the opposite side. Three carotidswith greater than 70% stenosis were detected in threedifferent patients.

Using calcium volumes of 0.01, 0.03, 0.06, 0.09 and0.12 cc as thresholds for evaluation for ‘‘significant’’stenoses, the sensitivity, specificity, positive predictivevalue, and negative predictive value were calculated(Table 1). A ROC curve was also generated using thesensitivity and specificity (Fig. 2). The sensitivity was100% for both sides when using a minimum threshold of0.01 cc, but the specificity was only 56%. The specificitywas 93% for both sides when using 0.12 cc as athreshold, but the sensitivity was only 56%. Thresholdsof 0.06 cc (sensitivity 88%, specificity 87%) and 0.03 cc(sensitivity 94%, specificity 76%) generated the bestcombination of sensitivity and specificity.

Using the data from all 117 carotid bifurcations, wecomputed the correlation coefficient (r) between cal-cium volume and percent stenosis. This correlationwas moderate to strong with r = 0.65 for all carotidstogether (r = 0.72 for right side only, 0.62 for leftonly, P<0.0001 in all cases). The scatter plot in Fig. 3demonstrates the relationship between percent stenosisand the corresponding measured calcium volume. Itwas noted that the on the scatter plot the calciumvolume appeared to increase faster than the degree ofstenosis, and hence calculations were performed forthe correlation coefficient r for various permutations ofthe calcium volumes (square root, cube root, square,etc. of the calcium volume) to find the manipulation ofthe calcium volume that would produce the greatestcorrelation with the percent stenosis. We found thatthe square root of the measured calcium volume had amore linear relationship and hence a stronger corre-lation to percent stenosis (r = 0.77 for all carotidstogether, r = 0.79 for the right carotids, r = 0.75 forthe left carotids, P<0.0001 in all cases) than the ac-tual calcium volume itself. The scatter plot with linearregression curve in Fig. 4 shows the relationship be-tween the percent stenosis and this square root of thecalcium volume. We note that the use of the squareroot of the calcium volume as a threshold to detect forstenosis does not change the sensitivity or specificity ofour test when using the square root of the calciumvolume and thresholds of 0.01–0.12 cc. We also notethat in the two scatter plots (Fig. 3, 4), there weremany carotids (59 total) where there was no measur-able calcium or percent stenosis (calcium volume andpercent stenosis both = 0). Thus for clarity, we ex-panded out the zero point on both axes so that alldata points could be visualized.

Table 1 Sensitivities and specificities at the five calcium volumethresholds that were used to evaluate for a ‘‘significant’’ stenosis.PPV positive predictive value, NPV negative predictive value

0.01 cc 0.03 cc 0.06 cc 0.09 cc 0.12 cc

Sensitivity (%) 100.0 93.8 87.5 75.0 56.3Specificity (%) 55.6 75.6 86.7 91.1 93.3PPV (%) 44.4 57.7 70.0 75.0 75.0NPV (%) 100.0 97.1 95.1 91.1 85.7

A stenosis of 40% or greater was selected as positive for ‘‘signifi-cant’’ stenosis. Thresholds of 0.03 cc and 0.06 cc have the highestcombination of sensitivity and specificity. Statistically, the squareroots of these five thresholds provide the same sensitivity, speci-ficity, PPV and NPV so a duplicate table was not included forpurposes of brevity

Fig. 2 The ROC curve is composed of five thresholds: 0.01, 0.03,0.06, 0.09, and 0.12 cc. The ROC curve illustrates that the calciumvolume thresholds of 0.03 and 0.06 cc are relatively sensitive andspecific for detecting a >40% ICA stenosis

5

Discussion

Our paper was designed to test the sensitivity/specificityand feasibility of carotid bifurcation calcium volumemeasurements in detecting ‘‘significant’’ ICA stenosis,

and to correlate those volume measurements with thedegree of stenosis. Regarding the use of calcium mea-surements, the diagnostic use of CT for calcified plaquemeasurements was initially developed for the detectionand screening of patients with clinical coronary arterydisease, and significant differences in calcium ‘‘score’’

Fig. 4 Scatter plot with linearregression curve demonstratesthe strong correlation (r =0.77) between percent ICA ste-nosis and square root of thecalcium volume for all 117carotid bifurcations. Note that59 carotids were randomized atthe zero point (where carotidbifurcation volume = 0 andpercent stenosis = 0) to betterillustrate this number of pa-tients and to prevent thesevalues from visually appearingas one point at the x= 0, y= 0axis

Fig. 3 Scatter plot demon-strates correlation between per-cent ICA stenosis and calciumvolume (in cc) of the carotidbifurcation

6

(based on the number, areas and peak CT Houndsfieldunits of the calcific lesions) were noted between thosewith versus those without clinical coronary artery dis-ease [15]. Ultrafast CT had already been shown to bemore sensitive and specific than fluoroscopy for detect-ing coronary calcium [16]. The sensitivity and specificityof helical CT in detecting significant/obstructive coro-nary artery disease (defined as greater than 50% steno-sis) when considering plaque density measurements hasbeen reported previously as 88 and 52%, respectively [5].Hence, we set out to evaluate if measurements of theamount of bifurcation calcium (in this case measured asa volume) would correlate with ICA stenoses, althoughwe could not use an identical system to that in the cor-onaries due to the inherent criteria (e.g. multiple vesselevaluation factors into coronary scoring).

In our study of carotid calcium volumes, we obtainedsensitivities of 94 and 88% at 0.03 and 0.06 cc, respec-tively, and specificities of 76 and 87% for detecting>40% stenoses as determined by CTA. On the basis ofthis preliminary study, use of a calcium volume as athreshold to evaluate for stenosis appears as sensitiveand more specific than helical CT and EBCT coronaryscoring in detection of ‘‘significant’’ internal carotidluminal stenosis near the bifurcation. In the literature,there has also been a strong correlation between coro-nary calcium scoring positives on EBCT (using aHoundsfield unit threshold) and degree of coronarystenosis in the involved vessel on autopsy (Pearson’scorrelation coefficient r = 0.66) [8]. Our study found arelatively strong correlation (r = 0.65) between thecalcium volume and percent stenosis on CTA. Since wenoted that this relatively strong correlation was present,we sought to define a manipulation or equation of thecalcium volume that would better correlate with thepercent stenosis on CTA (although this would notchange the sensitivity and specificity of our test). Ourstatistician found that the use of the square root of thecalcium volume has an even stronger and more linearcorrelation (r = 0.77) with percent stenosis as measuredby CTA.

We chose CTA using NASCET-like criteria as asurrogate measure of stenosis over the usual ‘‘gold’’standard of conventional angiography. This choicepermitted our being able to simultaneously acquire CTcalcium volume and CTA stenosis measurements in thesame patient (who was being scanned for other pur-poses) in a single evaluation. The 1-mm collimationcapability of the multidetector CT scanner allowed thisdata to be retrospectively reconstructed from the diag-nostic neck CT data. Thus, a potential limitation is thatwe implemented NASCET-like criteria (originallydeveloped for angiography) by using CTA derivedmeasurements as the standard for our study. However,CTA measures of all degrees of stenosis have been pre-viously shown to correlate very strongly (r correlation

coefficient ranging from 0.892 to 0.987) with conven-tional catheter cervicocerebral angiography [11, 13, 17]and to have a stronger correlation than ultrasound(0.808) compared to conventional catheter cervicocere-bral angiography [17]. CTA has been reported as verysensitive (89%), specific (91%) and accurate (90%) indetecting if luminal stenosis is greater than 50% (versusDSA), but it is less sensitive in characterizing whetherthe stenosis is within the 50–69% range (65%) or the 70–99% range (73%) [15]. CTA has also been shown to beaccurate (relative to DSA) in characterizing the degreeof stenosis even in a ‘‘moderate’’ stenosis (defined as 30–50%) [18].

Our study was designed to evaluate if there is a cor-relation between percent stenosis on CTA and calciumburden, as measured on CT. We obtained our testpopulation from persons who were undergoing neck CTfor various clinical reasons, most of which were forevaluating head and neck neoplasms. These patientsoften have an inherently higher risk and prevalence ofatherosclerotic disease compared to the general popu-lation, so our test population is likely not truly repre-sentative of the general population. Therefore, this couldlead to a somewhat elevated rate of detection as well asan element of expectation bias within our studied sam-ple. However, these are patients at high risk for carotidstenosis, and early detection is the intent of this study,Hence, the goal of this study was to evaluate the utilityof this examination in those patients at a somewhathigher risk of atherosclertotic carotid disease. We arbi-trarily selected 50 years as the threshold age for our‘‘simulated’’ screening population. However, one couldargue that earlier ages in diabetics, hypertensive pa-tients, or patients with hypercholesterolemia could havebeen included. Also, there may be gender differences inage of onset of carotid disease or differences in age ofonset between ethnic groups, which was not accountedfor by our study.

We selected 40% or greater luminal stenosis to be‘‘significant’’, since ultrasound screening thresholds varywidely and some studies use 40% [19–21] as the lowerthreshold to evaluate for stenosis while others use 50%[22, 23] to evaluate for a significant stenosis. Since ahemodynamically significant (i.e. flow-limiting) stenosishas traditionally been considered to be 50% or greater,we decided on a threshold of 40% for the purposes ofpotentially identify the patients at risk before their le-sions become flow-limiting, since one of the primarypurposes of our study was directed towards the possi-bility of using the calcium volume measurement as a toolfor early detection of carotid disease.

A single observer performed the initial calculations ofboth stenosis and calcium burden. Hence, interoperatorvariability was not evaluated for with regard to stenosismeasurement, although the calcium volume measure-ment was relatively automated once the volume was

7

sculpted, since the window and level were manipulatedwithin a relatively narrow range of Hounsfield units. Wedid not notice a significant difference in the calciumvolume measured when manipulating the window andlevel within that range. However, this is a potentialsource of variability between observers.

We obtained the carotid calcium volumes fromstudies in which intravenous contrast medium was uti-lized, thus obligating the observer to segment the cal-cium from intraluminal contrast via windowing/levelmanipulation. Calcified plaque was easy to differentiatefrom intravascular contrast enhancement, but occa-sionally differentiating noncalcified plaque from vascu-lar enhancement was slightly more time consuming.Therefore, care was taken to accurately measure thestenosis in cases of noncalcified plaque (appearance ofnoncalcified combined with calcified plaque is demon-strated in Fig. 1f). Since the CTAs of all of the patientsin our study were reconstructed from routine neck CTswith intravenous contrast, noncontrast volume mea-surements were not obtained.

In theory, greater sensitivity for detection of carotidcalcium would be attained with noncontrast CT carotidcalcium volume measurements, since a single thresholdof calcium plaque density could be implemented (in amanner similar to the Hounsfield density threshold incoronary scoring), which could be performed even by atechnician. This volume analysis measurement couldalso potentially become automated by the developmentof computer software since calcified plaque would standout more readily from non-enhancing ICA vasculature(and it would not be necessary to visually sculpt a vol-ume as in our study), which was not evaluated here. Anon-contrast measurement of carotid calcium volume byCT would also be much faster to review (compared tothe volume measurements obtained from our contrastenhanced studies), as much of our time (2–5 min perside) was spent on manipulating the proper plane inwhich to measure the luminal diameter for stenosismeasurement and on segmenting vascular enhancementfrom calcified plaque. This non-contrast measurement ofcalcium volumes could serve in a similar fashion tocoronary calcium scoring in detecting ICA stenosis andcould even be an adjunct to coronary scoring, sincepatients with atherosclerotic coronary artery diseasehave been shown to have more rapid progression ofextracranial carotid artery disease [24].

Another possibility would be the use of the non-contrast measurement of calcium volumes as an adjunct

with carotid ultrasound as a screening tool. This couldbe derived from the observation that, as with carotidultrasound (generally in the range of 90–95% negativepredictive value [25, 26]), the measurements of calciumvolumes presented here show relatively high negativepredictive values (97% at 0.03 cc and 95% at 0.06 cc),which is considered one of the critical criteria of a po-tential screening method.

Another potential limitation of our study is thatcommon carotid artery (CCA), external carotid artery(ECA), and internal carotid artery calcium were in-cluded together in the total volume of calcium measured.Therefore, this volume is not derived solely from thecalcified plaque of the ICA, but is actually representativeof the overall carotid bifurcation calcified plaque bur-den. Hence, theoretically a patient could have a highvolume of calcium in the CCA (or even ECA) withoutsignificant calcified plaque burden or stenosis of theICA. This distribution of calcification is uncommon(and was not present in our study), since CCA stenosisor occlusion usually involves the ICA as well, but thisuncommon variant is a well-described phenomenon inthe literature [27, 28]. Theoretically, in a patient with thispathologic variant, a carotid bifurcation calcium volumeobtained by our method may detect a ‘‘significant’’ ste-nosis, but would actually represent a stenosis of theCCA rather than the intended ICA.

In conclusion, on this preliminary correlation study,measurement of the carotid bifurcation calcium volume(or square root of the calcium volume) demonstrates arelatively strong correlation with percent stenosis on CTangiography in a test population over 50 years of age. Inthis study, using carotid bifurcation calcium volumethresholds (measured via multidetector CT) appeared tobe relatively sensitive and specific and have a high neg-ative predictive value for detection of a greater than40% stenosis of the ICA below the skull base.

Future studies may be warranted that would correlatethe noncontrast carotid bifurcation calcium volume andthe degree of stenosis on ultrasound, CTA, MRA, orconventional catheter angiography. Theoretically, a‘‘score’’ derived from the noncontrast carotid bifurca-tion exam could be obtained (whether the calcium vol-ume itself or a numerical manipulation of such) whichwould be able to identify patients with atheroscleroticdisease at high risk for developing flow limiting lesionsof the ICA. Hence, the potential would exist for futureuse of these measurements as a possible non-invasivescreening tool.

8

References

1. Robinson JG, Leon AS (1994) Theprevention of cardiovascular disease.Emphasis on secondary prevention.Med Clin N Am 78:69–98

2. Alan S, Ulgen MS, Ozturk O, Alan B,Ozdemir L, Toprak N (2003) Relationbetween coronary artery disease, riskfactors and intima-media thickness ofcarotid artery, arterial distensibility,and stiffness index. Angiology 54:261–267

3. Simon A, Giral P, Levenson J (1995)Extracoronary atherosclerotic plaque atmultiple sites and total coronary calci-fication deposit in asymptomatic men.Association with coronary risk profile.Circulation 92:1414–1421

4. Crouse JR, Toole JF, McKinney WM,Dignan MB, Howard G, Kahl FR,McMahan MR, Harpold GH (1987)Risk factors for extracranial carotidartery atherosclerosis. Stroke 18:990–996

5. Broderick LS, Shemesh J, Wilensky RLet al. (1996) Measurement of coronaryartery calcium with dual-slice helical CTcompared with coronary angiography:evaluation of CT scoring methods, in-terobserver variations, and reproduc-ibility. Am J Roentgenol 167:439–444

6. Woodcock RJ Jr, Goldstein JH, Kall-mes DF, Cloft HJ, Phillips CD (1999)Angiographic correlation of CT calcifi-cation in the carotid siphon. AmJ Neuroradiol 20:495–499

7. Rumberger JA, Sheedy PF, Breen JF,Schwartz RS (1995) Coronary calcium,as determined by electron beam com-puted tomography, and coronary dis-ease on arteriogram. Circulation91:1363–1367

8. Simons DB, Schwartz RS, EdwardsWD, Sheedy PF, Breen JF, RumbergerJA (1992) Noninvasive definition ofanatomic coronary artery disease byultrafast computed tomographic scan-ning: a quantitative pathologic com-parison study. J Am Coll Cardiol20:1118–1126

9. Nederkoorn PJ, van der Graaf Y,Hunink MG (2003) Duplex ultrasoundand magnetic resonance angiographycompared with digital subtractionangiography in carotid artery stenosis: asystematic review (Epub 2003 Apr 10).Stroke 34(5):1324–1332

10. Johnston DC, Goldstein LB (2001)Clinical carotid endarterectomy deci-sion making: noninvasive vascularimaging versus angiography. Neurology56:1009–1015

11. Cumming MJ, Morrow IM (1995)Carotid artery stenosis: a prospectivecomparison of CT angiography andconventional angiography. AmJ Roentgenol 165:233–235

12. Cinat M, Lane CT, Pham H, Lee A,Wilson SE, Gordon I (1998) Helical CTangiography in the preoperative evalu-ation of carotid artery stenosis. J VascSurg 28:290–300

13. Sameshima T, Futami S, Morita Y,Yokogami K, Miyahara S, SameshimaY, Goya T, Wakisaka S (1999) Clinicalusefulness of and problems with three-dimensional CT angiography for theevaluation of arteriosclerotic stenosis ofthe carotid artery: comparison withconventional angiography, MRA, andultrasound sonography. Surg Neurol51:301–309

14. Beneficial effect of carotid endarterec-tomy in symptomatic patients with highgrade carotid stenosis (1991) NorthAmerican Symptomatic Carotid End-arterectomy Trial collaborators. N EnglJ Med 325:445–53

15. Janowitz WR, Agatston AS, ViamonteM Jr (1991) Comparison of serialquantitative evaluation of calcified cor-onary artery plaque by ultrafast com-puted tomography in persons with andwithout obstructive coronary arterydisease. Am J Cardiol 68:1–6

16. Agatston AS, Janowitz WR, HildnerFJ, Zusmer NR, Viamonte M Jr, Detr-ano R (1990) Quantification of coro-nary artery calcium using ultrafastcomputed tomography. J Am CollCardiol 15:827–832

17. Anderson GB, Ashforth R, Steinke DE,Ferdinandy R, Findlay JM (2000) CTangiography for the detection andcharacterization of carotid arterybifurcation disease. Stroke 31:2168–2174

18. Binaghi S, Maeder P, Uske A, MeuwlyJY, Devuyst G, Meuli RA (2001) Three-dimensional computed tomographyangiography and magnetic resonanceangiography of carotid bifurcation ste-nosis. Eur Neurol 46:25–34

19. White JE, Russell WL, Greer MS,Whittle MT (1994) Efficacy of screeningMR angiography and Doppler ultraso-nography in the evaluation of carotidartery stenosis. Am Surg 60:340–348

20. Muller M, Ciccotti P, Reiche W, HagenT (2001) Comparison of color-flowDoppler scanning, power Dopplerscanning, and frequency shift forassessment of carotid artery stenosis.J Vasc Surg 34:1090–1095

21. Arbeille P, Desombre C, Aesh B, Phil-ippot M, Lapierre F (1995) Quantifica-tion and assessment of carotid arterylesions: degree of stenosis and plaquevolume. J Clin Ultrasound 23:113–124

22. Zwiebel WJ, Austin CW, Sackett JF,Strother CM (1983) Correlation ofhigh-resolution, B-mode and continu-ous-wave Doppler sonography witharteriography in the diagnosis of carotidstenosis. Radiology 149:523–532

23. Blackshear WM Jr, Lamb SL, KolliparaVS, Anderson JD, Murtagh FR, ShahCP, Farber MS (1984) Correlation ofhemodynamically significant internalcarotid stenosis with pulsed Dopplerfrequency analysis. Ann Surg 199:475–481

24. Crouse JR 3rd, Tang R, Espeland MA,Terry JG, Morgan T, Mercuri M (2002)Associations of extracranial carotidatherosclerosis progression with coro-nary status and risk factors in patientswith and without coronary artery dis-ease. Circulation 106:2061–2066

25. Fischer GG, Anderson DC, Farber R,Lebow S (1985) Prediction of carotiddisease by ultrasound and digital sub-traction angiography. Arch Neurol42:224–227

26. Mittl RL Jr, Broderick M, Carpenter JPet al. (1994) Blinded-reader comparisonof magnetic resonance angiography andduplex ultrasonography for carotid ar-tery bifurcation stenosis. Stroke 25:4–10

27. Keller HM, Valavanis A, Imhof HG,Turina M (1984) Patency of externaland internal carotid artery in the pres-ence of an occluded common carotidartery: noninvasive evaluation withcombined cerebrovascular Dopplerexamination and sequential computertomography. Stroke 15:149–157

28. Riles TS, Imparato AM, Posner MP,Eikelboom BC (1984) Common carotidocclusion. Assessment of the distal ves-sels. Ann Surg 199:363–366

9