Carotid bifurcation calcium and correlation with percent ...Document Transcript
Neuroradiology (2005) 47: 1–9DOI 10.1007/s00234-004-1301-4 DIAGNOSTIC NEURORADIOLOGYAlexander M. McKinneySean O. Casey Carotid bifurcation calcium and correlationMehmet Teksam with percent stenosis of the internal carotidLeandro T. LucatoMaurice Smith artery on CT angiographyCharles L. TruwitStephen KieﬀerReceived: 30 December 2003 Abstract The aim of this paper was calcium. We found a relatively strongAccepted: 4 September 2004 to determine the correlation between correlation between percent stenosisPublished online: 14 January 2005 calcium burden (expressed as a vol- and the calcium volume (Pearson’s rÓ Springer-Verlag 2005 ume) and extent of stenosis of the = 0.65, P<0.0001). We also found origin of the internal carotid artery that there was an even stronger cor- (ICA) by CT angiography (CTA). relation between the square root of Previous studies have shown that the calcium volume and the percent calciﬁcation in the coronary arteries stenosis as measured by CTA (r= correlates with signiﬁcant vessel ste- 0.77, P<0.0001). Calcium volumes nosis, and severe calciﬁcation (mea- of 0.01, 0.03, 0.06, 0.09 and 0.12 cc sured by CT) in the carotid siphon were used as thresholds to evaluate correlates with signiﬁcant (greater for a ‘‘signiﬁcant’’ stenosis. A re- than 50% stenosis) as determined ceiver operating characteristic angiographically. Sixty-one patients (ROC) curve demonstrated that (age range 50–85 years) underwent thresholds of 0.06 cc (sensitivity CT of the neck with intravenous 88%, speciﬁcity 87%) and 0.03 cc administration of iodinated contrast (sensitivity 94%, speciﬁcity 76%) for a variety of conditions. Images generated the best combinations of were obtained with a helical multi- sensitivity and speciﬁcity. Hence, thisPresented at the 41st Annual Meeting of the detector array CT scanner and preliminary study demonstrates aAmerican Society of Neuroradiology, reviewed on a three-dimensional relatively strong relationship be-Washington D.C., 2003. Sean Casey, MD workstation. A single observer tween volume of calcium at the car-and Charles Truwit, MD are members of manipulated window and level to otid bifurcation in the neckthe Medical Advisory Board of Vital Ima- segment calciﬁed plaque from (measured by CT) and percent ste-ges (Plymouth, Minnesota), the companythat developed the Vitrea 2 software. vascular enhancement in order to nosis of the ICA below the skull base quantify vascular calcium volume (as measured by CTA). Use of cal-A. M. McKinney (&) Æ S. O. Casey (cc) in the region of the bifurcation of cium volume measurements as aM. Teksam Æ C. L. Truwit Æ S. KieﬀerDepartments of Radiology, University of the common carotid artery/ICA ori- threshold may be both sensitive andMinnesota Medical School, and Hennepin gin, and to measure the extent of speciﬁc for the detection of signiﬁ-County Medical Center, Minneapolis, ICA stenosis near the origin. A total cant ICA stenosis. The signiﬁcanceMN 55455, USA of 117 common carotid artery bifur- of the correlation between calciumE-mail: firstname.lastname@example.orgTel.: +1-612-2734092 cations were reviewed. A volume and ICA stenosis is thatFax: +1-612-6248495 ‘‘signiﬁcant’’ stenosis was deﬁned potentially a ‘‘score’’ can be obtained arbitrarily as >40% (to detect that will identify those at risk forL. T. LucatoClinics Hospital of the University of Sao lesions before they become hemody- high grade carotid stenosis.Paulo, Sao Paulo, Brazil namically signiﬁcant) of luminal diameter on CTA using NASCET- Keywords Calcium Æ CarotidM. SmithDepartment of Biomedical Engineering, like criteria. All ‘‘signiﬁcant’’ arteries Æ Calciﬁcation Æ ComputedJohns Hopkins University, Baltimore, stenoses (21 out of 117 carotid tomography (CT) Æ CT angiographyMD 21218, USA bifurcations) had measurable (CTA) Æ Carotid stenosis
2Introduction trast. Age (equal to or greater than 50 years) was the only signiﬁcant factor for inclusion in this study. Pa-Atherosclerotic disease in the coronary and carotid tients were not subselected for diabetes or hypertensionarteries is a signiﬁcant cause of morbidity and mortality or other risk factors known to accelerate atherosclerotic. Coronary and carotid artery atherosclerotic disease disease. These examinations were performed for variousshare similar risk factors and are often coexistent in the indications, mostly for the purpose of evaluating headat-risk population, especially diabetics, hypertensive and neck cancer. None of the examinations was per-patients, smokers, patients with hypercholesterolemia formed for suspected carotid occlusive disease. Thirty-and those of advanced age [2–4]. Calciﬁcation in the eight females and 23 males were included. One patientcoronary arteries has been correlated with signiﬁcant with a known large invasive neck malignancy was ex-vessel stenosis (deﬁned as greater than 50%) . Severe cluded on the side of carotid encasement (right side),calciﬁcation in the carotid siphon has also been associ- since it was diﬃcult to distinguish carotid lumenated with a greater than 50% carotid stenosis . Cor- throughout most of the neck. In four patients withonary calcium scoring has been implemented for early known carotid endarterectomy, the side of involvementdetection of stenoses in high risk or symptomatic pa- was also excluded on the side of surgery (two on thetients, and usually focuses on thresholds of density right, two on the left), since we considered the extent ofmeasurements obtained by helical CT or electron beam disease to be unrepresentative of the true or past calciumCT (EBCT) [5, 7, 8], although this technique has burden; however, we included the non-operative side inencountered controversy due to being sensitive (88%) all four cases. Hence, a total of ﬁve carotid bifurcationsbut not quite as speciﬁc (52%) . were excluded (out of 122 possible carotids in 61 pa- Ultrasound and MRA are commonly used non-inva- tients).sive methods to evaluate for carotid stenosis in an at-risk All patients were evaluated with post-contrast neckpopulation, and both modalities have been demonstrated CT studies, which were obtained on a quad detectorto be sensitive for the detection of carotid stenosis, al- array system (Siemens Volume Zoom; Forcheim, Ger-though ultrasound has been described as slightly less many). Scanning was performed using 120 kVp with aaccurate and speciﬁc than MRA, particularly in patients maximum tube current of 230 mA (adjustable eﬀectivewith less than 70% stenosis [9, 10]. CTA has a proven mA depending on patient body habitus). Scanning timestrong correlation with percent stenosis in evaluating for was approximately 15 s, and images were obtainedinternal carotid artery (ICA) stenosis near the carotid during a 15–20 s breath hold. Scans were performed in abifurcation, relative to the gold-standard conventional caudal-cranial direction from the level of the aortic archcatheter angiography [11–13]. However, CTA is not up to 1 cm above the level of the sella in order to includetypically used as the initial evaluation for carotid stenosis the Circle of Willis. Source images were displayed with adue to expense and risks associated with intravenous 20–24 cm ﬁeld of view depending on patient body hab-administration of iodinated contrast material. itus. A 22 gauge catheter or larger was used for injection The primary goal of this study was to test the of contrast medium in peripheral veins in either thehypothesis that a sample volume of calcium at the car- antecubital region, forearm, or hand, for a total ofotid bifurcation may correlate with percent luminal ste- 150 cc. Pre-intravenous contrast scans were not ob-nosis. CTA was used as the relative gold standard in our tained. Non-ionic contrast material (ioversol 320 mg/ml;study to evaluate the percent of luminal internal carotid Mallinckrodt, St Louis, Miss., USA) was administeredstenosis near the bifurcation. As a preliminary study to in the following fashion: 50 cc were injected manuallydetermine the feasibility of using carotid calcium volume prior to obtaining a scout topogram, and then scanto detect ‘‘signiﬁcant’’ stenoses, we studied a series of range was set. Within minutes after the manual injec-routine post-intravenous contrast neck CT examinations tion, the power injector was connected to the catheter,performed on a multidetector CT scanner. The data sets and contrast infusion started at a rate of 3 cc/s for theobtained from these scans allowed for simultaneous ﬁrst 70 cc and 2 cc/s for the last 30 cc. Delay betweencalcium volume measurements while also including the injection and scan was 30 s. If signiﬁcant dental amal-CT angiographic data for stenosis measurements. gam was noted on the scout topogram, power injection rate was 3 cc/s for the ﬁrst 40 cc, 1.5 cc/s for the next 35 cc, and after a 20 s delay (and reangling of scan planeMaterials and methods through the region of the mouth), the last 25 cc was power injected at a rate of 2.8 cc/s, 1 mm collimationSixty-one consecutive patients (age range 50–85 years, was used to produce 4 mm thick images for routinemean 63 years) were included in the study, all of whom clinical use. Raw data were then retrospectively recon-underwent clinically indicated neck CT with a helical structed to 1 mm thick sections at 0.5 mm intervals formultidetector array (MDA) CT scanner following a use in calcium volume measurement and CTA evalua-bolus of intravenous administration of iodinated con- tion of the carotid arteries.
3Fig. 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
4 The reconstructed raw data were then sent to a Vitrea the observer measured the cross-sectional luminal2 workstation (Vital Images, Plymouth, Minn., USA) diameter in millimeters at the narrowest locationfor evaluation of CTA and quantiﬁcation of total vas- (Fig. 1e) (even in the absence of signiﬁcant plaque) andcular calcium volume by a single observer. The observer measured the more normal/non-stenotic internal carotidwas able to view vessels in an inﬁnite number of multi- artery diameter at least 2 cm above the site of narrow-planar views, including oblique angulations. The ob- ing/plaque (Fig. 1f). Thus, an attempt was made to ad-server would ﬁrst evaluate using various windows and here to North American Symptomatic Carotidlevels for mass, trauma or other extravascular etiology Endarterectomy Trial (NASCET) criteria as closely asthat could alter visualization or measurement of the possible  when performing stenosis measurements.carotid bifurcation (Fig. 1a, b). Locations of narrowing in the ICA were evaluated at For the purpose of carotid bifurcation calcium bur- any level from the carotid bifurcation to the skull base;den measurement, a volume was manually ‘‘sculpted’’ common carotid artery narrowing and intracranial ICAfrom 2 above to 2 cm below the carotid bifurcation narrowing were not evaluated in this study. Images ofwhile taking care not to include thyroid cartilage, cricoid the ﬁnal calcium burden volume and measurements ofcartilage, or styloid process calciﬁcations (Fig. 1c). This luminal diameter were stored for future reference. A‘‘sculpting’’ consisted of drawing the area intended for single observer initially evaluated all studies, although ainclusion on serial axial images, so that other sources of second observer re-evaluated those patients with ‘‘sig-calciﬁcation/bone (e.g. thyroid/cricoid cartilage and niﬁcant’’ stenoses.vertebra) were outside of the drawn area and hence ex- For statistical analysis, the sensitivity and speciﬁcitycluded from the resulting sculpted volume. External as well as a receiver operating characteristic (ROC)carotid calciﬁcation near the bifurcation was not sepa- curve of the study data were obtained at ﬁve arbitraryrated from that in the internal carotid artery while volume thresholds of calcium burden: 0.01, 0.03, 0.06,sculpting. After sculpting, the window and level were 0.09, and 0.12 cc. We arbitrarily used 40% as ourthen manually altered to maximize the amount of visu- deﬁnition for ‘‘signiﬁcant stenosis’’ in order to detect aalized calcium (level 250–500 HU, window 10–20 HU) stenotic lesion before the stenosis reached 50% (whichwhile not including vascular contrast (Fig. 1d). This is commonly considered the threshold at which a ste-combination of sculpting and windowing would result in nosis is hemodynamically signiﬁcant). We considered avisualization of only the bright/dense bifurcation cal- positive test one in which the patient had a greatercium. Using a volume calculation function available on than 40% luminal stenosis in one or both of the ICAsthe workstation, plaque burden was calculated in cc. The in the presence of the calcium burden threshold atsoftware would automatically calculate the remaining which we were testing. A Pearson’s correlation coeﬃ-volume (in cc) of the visualized calcium by the press of a cient (r) was calculated to assess for degree of corre-button. The smallest plaque volume that can be calcu- lation between calcium volume and percent stenosis.lated by the software is 0.01 cc. The correlation coeﬃcient (r) can range from )1.0 to After excluding extraneous cause of carotid involve- 1.0, where )1.0 is a perfect negative (inverse) correla-ment (such as tumor encasement/invasion), bifurcation tion, 0.0 is no correlation, and 1.0 is a perfect positivevascular anatomy was evaluated by manipulating win- correlation. Typically, an r of 0.4–0.7 is considered adow/level (window 900–1200 HU, level 300–500 HU) to moderate correlation and 0.7–0.9 is a strong correla-view calciﬁed and non-calciﬁed plaque and measure tion. The statistical signiﬁcance of r was tested using ainternal carotid stenoses. Calciﬁed plaque was easily t-test. A statistician reviewed the data and these cal-distinguished from non-calciﬁed plaque by manually culations.and visually manipulating window/level. The singlereader was not blinded to results of calcium volumedetermination before measuring degree of stenosis. All Resultsmeasurements of stenosis were obtained on 1 mmthickness images to minimize volume averaging. To Vascular calciﬁcation was not diﬃcult to distinguishavoid overestimation of luminal diameter, the images and segment from luminal carotid contrast enhance-used to measure luminal diameter stenosis were always ment on the workstation. Total time to measure cal-obtained from an image plane perpendicular (90°) to the cium burden and stenosis was approximately 2–5 minaxis of the vessel. This often required reorientation of per side, depending largely on the tortuosity of thethe image plane. Vascular luminal contrast enhancement vessel and the degree of calciﬁcation. Thirty-ﬁve of thewas easily separated visually from calciﬁed plaque (if 61 patients in this study demonstrated measurablepresent) by manual manipulation of window/level set- carotid bifurcation calcium burden greater than 0 cctings. This manipulation was monitored closely to avoid (27 on the right, 30 on the left), 21 of whom hadaccidental inclusion of vascular enhancement in the bilateral calciﬁcation. We noted that the computercalcium volume. To evaluate for the presence of stenosis, truncated two cases of very tiny amounts of visualized
5calcium (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 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 calcium‘‘signiﬁcant’’ stenoses, the mean calcium burden was volume thresholds of 0.03 and 0.06 cc are relatively sensitive and0.215 cc (SD 0.204). All ‘‘signiﬁcant’’ stenoses dem- speciﬁc for detecting a >40% ICA stenosisonstrated 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, Using the data from all 117 carotid bifurcations, weboth of which had previously undergone prior carotid computed the correlation coeﬃcient (r) between cal-endarterectomy on the opposite side. Three carotids cium volume and percent stenosis. This correlationwith greater than 70% stenosis were detected in three was moderate to strong with r = 0.65 for all carotidsdiﬀerent patients. together (r = 0.72 for right side only, 0.62 for left Using calcium volumes of 0.01, 0.03, 0.06, 0.09 and only, P<0.0001 in all cases). The scatter plot in Fig. 30.12 cc as thresholds for evaluation for ‘‘signiﬁcant’’ demonstrates the relationship between percent stenosisstenoses, the sensitivity, speciﬁcity, positive predictive and the corresponding measured calcium volume. Itvalue, and negative predictive value were calculated was noted that the on the scatter plot the calcium(Table 1). A ROC curve was also generated using the volume appeared to increase faster than the degree ofsensitivity and speciﬁcity (Fig. 2). The sensitivity was stenosis, and hence calculations were performed for100% for both sides when using a minimum threshold of the correlation coeﬃcient r for various permutations of0.01 cc, but the speciﬁcity was only 56%. The speciﬁcity the calcium volumes (square root, cube root, square,was 93% for both sides when using 0.12 cc as a etc. of the calcium volume) to ﬁnd the manipulation ofthreshold, but the sensitivity was only 56%. Thresholds the calcium volume that would produce the greatestof 0.06 cc (sensitivity 88%, speciﬁcity 87%) and 0.03 cc correlation with the percent stenosis. We found that(sensitivity 94%, speciﬁcity 76%) generated the best the square root of the measured calcium volume had acombination of sensitivity and speciﬁcity. more linear relationship and hence a stronger corre- lation to percent stenosis (r = 0.77 for all carotids together, r = 0.79 for the right carotids, r = 0.75 for the left carotids, P<0.0001 in all cases) than the ac-Table 1 Sensitivities and speciﬁcities at the ﬁve calcium volume tual calcium volume itself. The scatter plot with linearthresholds that were used to evaluate for a ‘‘signiﬁcant’’ stenosis. regression curve in Fig. 4 shows the relationship be-PPV positive predictive value, NPV negative predictive value tween the percent stenosis and this square root of the 0.01 cc 0.03 cc 0.06 cc 0.09 cc 0.12 cc calcium volume. We note that the use of the square root of the calcium volume as a threshold to detect forSensitivity (%) 100.0 93.8 87.5 75.0 56.3 stenosis does not change the sensitivity or speciﬁcity ofSpeciﬁcity (%) 55.6 75.6 86.7 91.1 93.3 our test when using the square root of the calciumPPV (%) 44.4 57.7 70.0 75.0 75.0NPV (%) 100.0 97.1 95.1 91.1 85.7 volume and thresholds of 0.01–0.12 cc. We also note that in the two scatter plots (Fig. 3, 4), there wereA stenosis of 40% or greater was selected as positive for ‘‘signiﬁ- many carotids (59 total) where there was no measur-cant’’ stenosis. Thresholds of 0.03 cc and 0.06 cc have the highest able calcium or percent stenosis (calcium volume andcombination of sensitivity and speciﬁcity. Statistically, the squareroots of these ﬁve thresholds provide the same sensitivity, speci- percent stenosis both = 0). Thus for clarity, we ex-ﬁcity, PPV and NPV so a duplicate table was not included for panded out the zero point on both axes so that allpurposes of brevity data points could be visualized.
6Fig. 3 Scatter plot demon-strates correlation between per-cent ICA stenosis and calciumvolume (in cc) of the carotidbifurcation and to correlate those volume measurements with theDiscussion degree of stenosis. Regarding the use of calcium mea- surements, the diagnostic use of CT for calciﬁed plaqueOur paper was designed to test the sensitivity/speciﬁcity measurements was initially developed for the detectionand feasibility of carotid bifurcation calcium volume and screening of patients with clinical coronary arterymeasurements in detecting ‘‘signiﬁcant’’ ICA stenosis, disease, 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
7(based on the number, areas and peak CT Houndsﬁeld coeﬃcient ranging from 0.892 to 0.987) with conven-units of the calciﬁc lesions) were noted between those tional catheter cervicocerebral angiography [11, 13, 17]with versus those without clinical coronary artery dis- and to have a stronger correlation than ultrasoundease . Ultrafast CT had already been shown to be (0.808) compared to conventional catheter cervicocere-more sensitive and speciﬁc than ﬂuoroscopy for detect- bral angiography . CTA has been reported as verying coronary calcium . The sensitivity and speciﬁcity sensitive (89%), speciﬁc (91%) and accurate (90%) inof helical CT in detecting signiﬁcant/obstructive coro- detecting if luminal stenosis is greater than 50% (versusnary artery disease (deﬁned as greater than 50% steno- DSA), but it is less sensitive in characterizing whethersis) when considering plaque density measurements has the stenosis is within the 50–69% range (65%) or the 70–been reported previously as 88 and 52%, respectively . 99% range (73%) . CTA has also been shown to beHence, we set out to evaluate if measurements of the accurate (relative to DSA) in characterizing the degreeamount of bifurcation calcium (in this case measured as of stenosis even in a ‘‘moderate’’ stenosis (deﬁned as 30–a volume) would correlate with ICA stenoses, although 50%) .we could not use an identical system to that in the cor- Our study was designed to evaluate if there is a cor-onaries due to the inherent criteria (e.g. multiple vessel relation between percent stenosis on CTA and calciumevaluation factors into coronary scoring). burden, as measured on CT. We obtained our test In our study of carotid calcium volumes, we obtained population from persons who were undergoing neck CTsensitivities of 94 and 88% at 0.03 and 0.06 cc, respec- for various clinical reasons, most of which were fortively, and speciﬁcities of 76 and 87% for detecting evaluating head and neck neoplasms. These patients>40% stenoses as determined by CTA. On the basis of often have an inherently higher risk and prevalence ofthis preliminary study, use of a calcium volume as a atherosclerotic disease compared to the general popu-threshold to evaluate for stenosis appears as sensitive lation, so our test population is likely not truly repre-and more speciﬁc than helical CT and EBCT coronary sentative of the general population. Therefore, this couldscoring in detection of ‘‘signiﬁcant’’ internal carotid lead to a somewhat elevated rate of detection as well asluminal stenosis near the bifurcation. In the literature, an element of expectation bias within our studied sam-there has also been a strong correlation between coro- ple. However, these are patients at high risk for carotidnary calcium scoring positives on EBCT (using a stenosis, and early detection is the intent of this study,Houndsﬁeld unit threshold) and degree of coronary Hence, the goal of this study was to evaluate the utilitystenosis in the involved vessel on autopsy (Pearson’s of this examination in those patients at a somewhatcorrelation coeﬃcient r = 0.66) . Our study found a higher risk of atherosclertotic carotid disease. We arbi-relatively strong correlation (r = 0.65) between the trarily selected 50 years as the threshold age for ourcalcium volume and percent stenosis on CTA. Since we ‘‘simulated’’ screening population. However, one couldnoted that this relatively strong correlation was present, argue that earlier ages in diabetics, hypertensive pa-we sought to deﬁne a manipulation or equation of the tients, or patients with hypercholesterolemia could havecalcium volume that would better correlate with the been included. Also, there may be gender diﬀerences inpercent stenosis on CTA (although this would not age of onset of carotid disease or diﬀerences in age ofchange the sensitivity and speciﬁcity of our test). Our onset between ethnic groups, which was not accountedstatistician found that the use of the square root of the for by our study.calcium volume has an even stronger and more linear We selected 40% or greater luminal stenosis to becorrelation (r = 0.77) with percent stenosis as measured ‘‘signiﬁcant’’, since ultrasound screening thresholds varyby CTA. widely and some studies use 40% [19–21] as the lower We chose CTA using NASCET-like criteria as a threshold to evaluate for stenosis while others use 50%surrogate measure of stenosis over the usual ‘‘gold’’ [22, 23] to evaluate for a signiﬁcant stenosis. Since astandard of conventional angiography. This choice hemodynamically signiﬁcant (i.e. ﬂow-limiting) stenosispermitted our being able to simultaneously acquire CT has traditionally been considered to be 50% or greater,calcium volume and CTA stenosis measurements in the we decided on a threshold of 40% for the purposes ofsame patient (who was being scanned for other pur- potentially identify the patients at risk before their le-poses) in a single evaluation. The 1-mm collimation sions become ﬂow-limiting, since one of the primarycapability of the multidetector CT scanner allowed this purposes of our study was directed towards the possi-data to be retrospectively reconstructed from the diag- bility of using the calcium volume measurement as a toolnostic neck CT data. Thus, a potential limitation is that for early detection of carotid disease.we implemented NASCET-like criteria (originally A single observer performed the initial calculations ofdeveloped for angiography) by using CTA derived both stenosis and calcium burden. Hence, interoperatormeasurements as the standard for our study. However, variability was not evaluated for with regard to stenosisCTA measures of all degrees of stenosis have been pre- measurement, although the calcium volume measure-viously shown to correlate very strongly (r correlation ment was relatively automated once the volume was
8sculpted, since the window and level were manipulated with carotid ultrasound as a screening tool. This couldwithin a relatively narrow range of Hounsﬁeld units. We be derived from the observation that, as with carotiddid not notice a signiﬁcant diﬀerence in the calcium ultrasound (generally in the range of 90–95% negativevolume measured when manipulating the window and predictive value [25, 26]), the measurements of calciumlevel within that range. However, this is a potential volumes presented here show relatively high negativesource of variability between observers. predictive values (97% at 0.03 cc and 95% at 0.06 cc), We obtained the carotid calcium volumes from which is considered one of the critical criteria of a po-studies in which intravenous contrast medium was uti- tential screening method.lized, thus obligating the observer to segment the cal- Another potential limitation of our study is thatcium from intraluminal contrast via windowing/level common carotid artery (CCA), external carotid arterymanipulation. Calciﬁed plaque was easy to diﬀerentiate (ECA), and internal carotid artery calcium were in-from intravascular contrast enhancement, but occa- cluded together in the total volume of calcium measured.sionally diﬀerentiating noncalciﬁed plaque from vascu- Therefore, this volume is not derived solely from thelar enhancement was slightly more time consuming. calciﬁed plaque of the ICA, but is actually representativeTherefore, care was taken to accurately measure the of the overall carotid bifurcation calciﬁed plaque bur-stenosis in cases of noncalciﬁed plaque (appearance of den. Hence, theoretically a patient could have a highnoncalciﬁed combined with calciﬁed plaque is demon- volume of calcium in the CCA (or even ECA) withoutstrated in Fig. 1f). Since the CTAs of all of the patients signiﬁcant calciﬁed plaque burden or stenosis of thein our study were reconstructed from routine neck CTs ICA. This distribution of calciﬁcation is uncommonwith intravenous contrast, noncontrast volume mea- (and was not present in our study), since CCA stenosissurements were not obtained. or occlusion usually involves the ICA as well, but this In theory, greater sensitivity for detection of carotid uncommon variant is a well-described phenomenon incalcium would be attained with noncontrast CT carotid the literature [27, 28]. Theoretically, in a patient with thiscalcium volume measurements, since a single threshold pathologic variant, a carotid bifurcation calcium volumeof calcium plaque density could be implemented (in a obtained by our method may detect a ‘‘signiﬁcant’’ ste-manner similar to the Hounsﬁeld density threshold in nosis, but would actually represent a stenosis of thecoronary scoring), which could be performed even by a CCA rather than the intended ICA.technician. This volume analysis measurement could In conclusion, on this preliminary correlation study,also potentially become automated by the development measurement of the carotid bifurcation calcium volumeof computer software since calciﬁed plaque would stand (or square root of the calcium volume) demonstrates aout more readily from non-enhancing ICA vasculature relatively strong correlation with percent stenosis on CT(and it would not be necessary to visually sculpt a vol- angiography in a test population over 50 years of age. Inume as in our study), which was not evaluated here. A this study, using carotid bifurcation calcium volumenon-contrast measurement of carotid calcium volume by thresholds (measured via multidetector CT) appeared toCT would also be much faster to review (compared to be relatively sensitive and speciﬁc and have a high neg-the volume measurements obtained from our contrast ative predictive value for detection of a greater thanenhanced studies), as much of our time (2–5 min per 40% stenosis of the ICA below the skull base.side) was spent on manipulating the proper plane in Future studies may be warranted that would correlatewhich to measure the luminal diameter for stenosis the noncontrast carotid bifurcation calcium volume andmeasurement and on segmenting vascular enhancement the degree of stenosis on ultrasound, CTA, MRA, orfrom calciﬁed plaque. This non-contrast measurement of conventional catheter angiography. Theoretically, acalcium volumes could serve in a similar fashion to ‘‘score’’ derived from the noncontrast carotid bifurca-coronary calcium scoring in detecting ICA stenosis and tion exam could be obtained (whether the calcium vol-could even be an adjunct to coronary scoring, since ume itself or a numerical manipulation of such) whichpatients with atherosclerotic coronary artery disease would be able to identify patients with atherosclerotichave been shown to have more rapid progression of disease at high risk for developing ﬂow limiting lesionsextracranial carotid artery disease . of the ICA. Hence, the potential would exist for future Another possibility would be the use of the non- use of these measurements as a possible non-invasivecontrast measurement of calcium volumes as an adjunct screening tool.
9References1. Robinson JG, Leon AS (1994) The 10. Johnston DC, Goldstein LB (2001) 20. Muller M, Ciccotti P, Reiche W, Hagen prevention of cardiovascular disease. Clinical carotid endarterectomy deci- T (2001) Comparison of color-ﬂow Emphasis on secondary prevention. sion making: noninvasive vascular Doppler scanning, power Doppler Med Clin N Am 78:69–98 imaging versus angiography. Neurology scanning, and frequency shift for2. Alan S, Ulgen MS, Ozturk O, Alan B, 56:1009–1015 assessment of carotid artery stenosis. Ozdemir L, Toprak N (2003) Relation 11. Cumming MJ, Morrow IM (1995) J Vasc Surg 34:1090–1095 between coronary artery disease, risk Carotid artery stenosis: a prospective 21. Arbeille P, Desombre C, Aesh B, Phil- factors and intima-media thickness of comparison of CT angiography and ippot M, Lapierre F (1995) Quantiﬁca- carotid artery, arterial distensibility, conventional angiography. Am tion and assessment of carotid artery and stiﬀness index. Angiology 54:261– J Roentgenol 165:233–235 lesions: degree of stenosis and plaque 267 12. Cinat M, Lane CT, Pham H, Lee A, volume. J Clin Ultrasound 23:113–1243. Simon A, Giral P, Levenson J (1995) Wilson SE, Gordon I (1998) Helical CT 22. Zwiebel WJ, Austin CW, Sackett JF, Extracoronary atherosclerotic plaque at angiography in the preoperative evalu- Strother CM (1983) Correlation of multiple sites and total coronary calci- ation of carotid artery stenosis. J Vasc high-resolution, B-mode and continu- ﬁcation deposit in asymptomatic men. Surg 28:290–300 ous-wave Doppler sonography with Association with coronary risk proﬁle. 13. Sameshima T, Futami S, Morita Y, arteriography in the diagnosis of carotid Circulation 92:1414–1421 Yokogami K, Miyahara S, Sameshima stenosis. Radiology 149:523–5324. Crouse JR, Toole JF, McKinney WM, Y, Goya T, Wakisaka S (1999) Clinical 23. Blackshear WM Jr, Lamb SL, Kollipara Dignan MB, Howard G, Kahl FR, usefulness of and problems with three- VS, Anderson JD, Murtagh FR, Shah McMahan MR, Harpold GH (1987) dimensional CT angiography for the CP, Farber MS (1984) Correlation of Risk factors for extracranial carotid evaluation of arteriosclerotic stenosis of hemodynamically signiﬁcant internal artery atherosclerosis. Stroke 18:990– the carotid artery: comparison with carotid stenosis with pulsed Doppler 996 conventional angiography, MRA, and frequency analysis. Ann Surg 199:475–5. Broderick LS, Shemesh J, Wilensky RL ultrasound sonography. Surg Neurol 481 et al. (1996) Measurement of coronary 51:301–309 24. Crouse JR 3rd, Tang R, Espeland MA, artery calcium with dual-slice helical CT 14. Beneﬁcial eﬀect of carotid endarterec- Terry JG, Morgan T, Mercuri M (2002) compared with coronary angiography: tomy in symptomatic patients with high Associations of extracranial carotid evaluation of CT scoring methods, in- grade carotid stenosis (1991) North atherosclerosis progression with coro- terobserver variations, and reproduc- American Symptomatic Carotid End- nary status and risk factors in patients ibility. Am J Roentgenol 167:439–444 arterectomy Trial collaborators. N Engl with and without coronary artery dis-6. Woodcock RJ Jr, Goldstein JH, Kall- J Med 325:445–53 ease. Circulation 106:2061–2066 mes DF, Cloft HJ, Phillips CD (1999) 15. Janowitz WR, Agatston AS, Viamonte 25. Fischer GG, Anderson DC, Farber R, Angiographic correlation of CT calciﬁ- M Jr (1991) Comparison of serial Lebow S (1985) Prediction of carotid cation in the carotid siphon. Am quantitative evaluation of calciﬁed cor- disease by ultrasound and digital sub- J Neuroradiol 20:495–499 onary artery plaque by ultrafast com- traction angiography. Arch Neurol7. Rumberger JA, Sheedy PF, Breen JF, puted tomography in persons with and 42:224–227 Schwartz RS (1995) Coronary calcium, without obstructive coronary artery 26. Mittl RL Jr, Broderick M, Carpenter JP as determined by electron beam com- disease. Am J Cardiol 68:1–6 et al. (1994) Blinded-reader comparison puted tomography, and coronary dis- 16. Agatston AS, Janowitz WR, Hildner of magnetic resonance angiography and ease on arteriogram. Circulation FJ, Zusmer NR, Viamonte M Jr, Detr- duplex ultrasonography for carotid ar- 91:1363–1367 ano R (1990) Quantiﬁcation of coro- tery bifurcation stenosis. Stroke 25:4–108. Simons DB, Schwartz RS, Edwards nary artery calcium using ultrafast 27. Keller HM, Valavanis A, Imhof HG, WD, Sheedy PF, Breen JF, Rumberger computed tomography. J Am Coll Turina M (1984) Patency of external JA (1992) Noninvasive deﬁnition of Cardiol 15:827–832 and internal carotid artery in the pres- anatomic coronary artery disease by 17. Anderson GB, Ashforth R, Steinke DE, ence of an occluded common carotid ultrafast computed tomographic scan- Ferdinandy R, Findlay JM (2000) CT artery: noninvasive evaluation with ning: a quantitative pathologic com- angiography for the detection and combined cerebrovascular Doppler parison study. J Am Coll Cardiol characterization of carotid artery examination and sequential computer 20:1118–1126 bifurcation disease. Stroke 31:2168– tomography. Stroke 15:149–1579. Nederkoorn PJ, van der Graaf Y, 2174 28. Riles TS, Imparato AM, Posner MP, Hunink MG (2003) Duplex ultrasound 18. Binaghi S, Maeder P, Uske A, Meuwly Eikelboom BC (1984) Common carotid and magnetic resonance angiography JY, Devuyst G, Meuli RA (2001) Three- occlusion. Assessment of the distal ves- compared with digital subtraction dimensional computed tomography sels. Ann Surg 199:363–366 angiography in carotid artery stenosis: a angiography and magnetic resonance systematic review (Epub 2003 Apr 10). angiography of carotid bifurcation ste- Stroke 34(5):1324–1332 nosis. Eur Neurol 46:25–34 19. White JE, Russell WL, Greer MS, Whittle MT (1994) Eﬃcacy of screening MR angiography and Doppler ultraso- nography in the evaluation of carotid artery stenosis. Am Surg 60:340–348