Comparison angiography, duplex angiography internal · digital subtraction angiography was-0-7...

1Journal ofNeurology, Neurosurgery, and Psychiatry 1994;57:1466-1478

Comparison of magnetic resonance angiography,duplex ultrasound, and digital subtractionangiography in assessment of extracranial internalcarotid artery stenosis

G R Young, P R D Humphrey, M D M Shaw, T E Nixon, E T S Smith

The Walton Centre forNeurology andNeurosurgery, RiceLane, Liverpool, UKG R YoungP R D HumphreyM D M ShawT E NixonE T S SmithCorrespondence to:Dr G R Young, The WaltonCentre for Neurology andNeurosurgery, Rice Lane,Liverpool L9 1AE, UK.Received 15 November 1993and in revised form21 March 1994.Accepted 1 July 1994

AbstractThe results of a prospective study com-paring ultrasound, intra-arterial digitalsubtraction angiography, and magneticresonance angiography in the assessmentof the degree of extracranial internalcarotid artery stenosis are reported inpatients with symptoms of recent carotidterritory ischaemia. A total of 70 patientsand 137 vessels were examined by allthree techniques. The results obtained byeach technique were reported blind. Themean difference (SD) for the comparisonof magnetic resonance angiography anddigital subtraction angiography was -0-7(14)%, for ultrasound and digital sub-traction angiography 3-1 (15)%, and formagnetic resonance angiography andultrasound -3-8 (15)%. The level ofagreement was greater for the moretightly stenosed vessels. With theassumption that the results of the digitalsubtraction angiogram reflect the truesituation, the sensitivity and specificity inthe detection of >30% stenoses were 93%and 82% with ultrasound and 89% and82% with magnetic resonance angiogra-phy; for stenoses >,70% 93% and 92%with ultrasound and 90% and 95% withmagnetic resonance angiography; and forstenoses of 70-99% 89% and 93% withultrasound and 86% and 93% with mag-netic resonance angiography. For occlu-sion the values were 93% and 99% withultrasound and 80% and 99% with mag-netic resonance angiography. Increasedsensitivity and specificity were obtainedwhen analysis was confined to those ves-sels in which ultrasound and magneticresonance angiography were in agree-ment over classification. It is thus possi-ble to accurately categorise the degree ofstenosis of the extracranial internalcarotid artery from a combination ofultrasound and magnetic resonanceangiography. The adoption ofthis combi-nation for the investigation of patientsbefore carotid endarterectomy removesthe risk associated with conventionalangiography and represents an importantadvance in the management of carotidstenosis.

(3 Neurol Neurosurg Psychiatry 1994;57:1466-1478)

The European Carotid Surgery Trial (ECST)and the North American SymptomaticCarotid Endarterectomy Trial (NASCET)

have shown the benefit of carotid endarterec-tomy over current best medical care forpatients with high grade stenosis of thesymptomatic internal carotid artery.' 2 Insuch patients, after carotid endarterectomy,the risk of disabling stroke or death is reducedby about 75%. This benefit exists onlyif the risk of stroke or death during investiga-tion and operation is low. If this riskexceeds 10%, the value of operation ismarginal.'

Adequately displaying the appropriatecarotid artery and accurately defining anystenosis present therefore plays a vital part inthe management of patients with presumedischaemic events in the territory of the carotidartery. The two techniques routinelyemployed for this purpose are duplex ultra-sound examination and x ray contrast angio-graphy. Whereas some centres routinelyperform carotid endarterectomy on the basisof results from duplex ultrasound alone,3-6most surgeons in the United Kingdom requireangiographic demonstration of the carotidbifurcation, by either conventional cerebralangiography or intra-arterial digital subtrac-tion angiography, before proceeding to oper-ate.7 Unfortunately, for this group of patientsthese invasive techniques carry a risk of causingstroke of between 1% and 4%.8 This riskseems to be greater for those patients in whomthere is appreciable atherosclerotic narrowingof the carotid artery. Our own series shows a2% risk of serious stroke for patients withmore than 30% stenosis of the carotid arteryundergoing selective intra-arterial digital sub-traction angiography.9 The risks of angiogra-phy should be added to those of operationwhen considering patients for endarterec-tomy. Therefore, while striving to reduce sur-gical risks, every effort must also be made toreduce the risks of angiography during thepreoperative assessment of these patients.Magnetic resonance angiography is a newtechnique that can produce angiograms non-invasively (fig 1). In a prospective study, weset out to measure the level of agreementbetween magnetic resonance angiography,intra-arterial digital subtraction angiography,and duplex ultrasound in determining thedegree of stenosis of the internal carotidartery, at or around the carotid bifurcation, inpatients under consideration for prophylacticcarotid endarterectomy.

MethodsThe study involved consecutive patientsreferred to the Walton Centre for Neurology

1466

on March 27, 2021 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.57.12.1466 on 1 D

ecember 1994. D

ownloaded from

http://jnnp.bmj.com/

Comparison of magnetic resonance angiography, duplex ultrasound, and digital subtraction angiography

A --Figure 1 (A) Intra-arterial digital subtraction angiogram showing ulcerated, irregular stenosis of distal common and proximal internal carotid arterywith (B) corresponding two dimensional, and (C) three dimensional magnetic resonance angiogram.

and Neurosurgery for angiography fromMarch 1992 to January 1993. Patients initiallyunderwent clinical evaluation and duplexultrasound examination at an outpatient cere-brovascular clinic. Those with symptoms ofrecent carotid territory ischaemic events andduplex ultrasound evidence of greater than30% stenosis of the appropriate carotid arterywere identified. Those wishing to proceedwith further investigations were admittedwithin two to three weeks for intra-arterialdigital subtraction angiography. Informedconsent was obtained from patients agreeingto undergo repeat duplex ultrasound exami-nation and magnetic resonance angiography,and all three techniques were then performedduring the same hospital admission.

ULTRASOUND TECHNIQUEUltrasound was performed in all cases by thesame experienced operator (PH), using bothduplex ultrasound examination of the neck(Diasonics C400) and continuous wave probeexamination of the neck and orbit(Krahnzbuhler 4 and 8MHz probe system).The final estimate of percentage stenosis wasbased on a combination of Doppler flow sig-nal and grey scale image with criteria previ-ously described.10

DIGITAL SUBTRACTION ANGIOGRAPHYAngiograms were performed under localanaesthetic with selective catheterisation of

the common carotid artery via the femoralartery in all except four cases, in which archangiograms were obtained. Intra-arterial digitalsubtraction angiograms were obtained with aSiemens digital subtraction unit (Siemens,Angioskop Digitron 2) except in one casewhere conventional "cut-film" angiogramswere obtained with the same equipment.Whenever possible three views-namely,anteroposterior, oblique, and lateral-wereobtained. In some cases, if the symptomaticcarotid artery was occluded, the contralateralcarotid artery was not examined.

MAGNETIC RESONANCE ANGIOGRAPHYMagnetic resonance angiography was per-formed on a 1-5 T Siemens Magnetom scan-ner. In one case the patient was positioned ina circularly polarised transmit and receiveSiemens head coil. In all other cases aHelmholtz, receive only, Siemens neck coilwas used. A short localising sequence todetermine the position of the carotid bifurca-tion and comprising 14 sagittal imagesthrough the neck was obtained in all cases.The most suitable image displaying the posi-tion of the carotid bifurcation was selectedand used to prescribe the other sequences. Allbut one patient underwent an axial twodimensional time of flight examination of thecarotid bifurcation covering a 100 mm regionfrom the distal common carotid artery to thebase of the skull, with 55 axial sections

1467



j.com/

J Neurol N


ecember 1994. D

ownloaded from


Young, Humphrey, Shaw, Nixon, Smith

(sequence settings; fast low angle shot(FLASH), repetition time 30 ms, echo time 9ms, flip angle 350, 23 cm field of view, 192 x

256 matrix, with travelling superior saturationband). Most patients, including the patient inwhom the two dimensional sequence was notperformed, also underwent an axial threedimensional time of flight examination cover-

ing a region 52 to 60 mm wide around thecarotid bifurcation (sequence settings; fastimaging with steady state precession (FISP),repetition time 36 ms, echo time 7 ms, flipangle 150, 64 partitions, 16 cm field of view,192 x 256 matrix). Postprocessing of theimages obtained was carried out with a maxi-mum intensity projection algorithm. Eachcarotid bifurcation was processed separatelyand wherever possible the vertebral arterieswere excluded from the processed region.Twelve views at 150 increments from 00 to1650 around each carotid bifurcation were

obtained. The time required for a typicalexamination comprising localising sequence

and two and three dimensional time of flightsequences was about 30 minutes, and thetime required for postprocessing the imageswas about 10 minutes.The duplex ultrasound examinations were

performed and interpreted by the same inves-tigator (PH) who was blinded to the results ofthe other investigations. The x ray and mag-

netic resonance angiograms were reportedindependently by one of two consultantneuroradiologists (TN or TS) experienced ininterpreting carotid angiograms and blindedfrom the results of the other investigations. Atthe end of the study the radiologists reviewedthe magnetic resonance and digital subtrac-tion angiograms a second time, so that an

assessment of between and within observervariation in reporting could be made. The twoand three dimensional magnetic resonance

angiograms were viewed together to give a

single result for magnetic resonance angio-graphy. The original axial images making upthe magnetic resonance angiograms were alsoavailable at the time of reporting, if requested.The digital subtraction and magnetic reso-

nance angiograms were then assessed fromcaliper measurements on two separate occa-

sions by the same investigator (GY).Mechanical calipers reading to 0 1 mm were

used for the digital subtraction films and elec-tronic calipers at the console for the magneticresonance angiograms. Measurements foreach technique were made at separate sittings,with a two week interval to reduce the likeli-hood of measurements by one method influ-encing those by the other. The diameter of theminimum residual lumen of the internalcarotid artery was used as numerator and thediameter of the common carotid artery lumenjust below the bifurcation as denominator togive the percentage residual lumen fromwhich the percentage stenosis was calculatedby subtraction from 100. A vessel showing a

"signal gap" by magnetic resonance angiogra-phy was arbitrarily assigned a result of 90%stenosis by caliper, as the residual lumencould not be measured in these circum-

stances. For the caliper analysis only, themagnetic resonance viewing angles were lim-ited to those obtained at digital subtractionangiography. This was done because magneticresonance angiography provides many moreviews of the bifurcation than are routinelyobtainable by digital subtraction angiography.Magnetic resonance angiography could there-fore report a tighter stenosis because of a"superior" viewing angle to digital subtractionangiography, rather than because of anyinherent difference between the representa-tion of the stenosis by the different tech-niques. Caliper measurements from magneticresonance angiograms acquired by the threedimensional technique were used for analysiswhere possible, with the two dimensionalangiograms assessed for those patients under-going only this examination.

MEASUREMENT OF THE DEGREE OF STENOSISWhen measuring the degree of stenosis of thecarotid artery, the percentage reduction inluminal diameter at the point of maximumstenosis is usually quoted. Often this figure isestimated by simple inspection of theangiogram by a radiologist. Strictly the valueshould be calculated by measuring the diame-ter of the original vessel lumen at the point ofmaximal stenosis, subtracting the diameter ofthe residual lumen at the same point, andexpressing the result as a percentage of theoriginal lumen to give the percentage stenosis.Problems arise because on a conventionalangiogram it is not possible to directly mea-sure the original lumen diameter. One solu-tion is to mark on the angiogram theestimated position of the original lumen andmake the appropriate measurements, asdetailed. This is the method used in theECST and has the obvious drawback that theresult is based on a subjective estimate ofwhere the vessel wall would be. TheNASCET used an alternative approachwhereby the lumen diameter of the distalinternal carotid artery, an area not oftenaffected by atheroma, was used in place of thetheoretical original lumen diameter. This hasthe advantage of being an objective measure-ment, presumably with less observer variation,but the disadvantage that the result obtainedis no longer the percentage reduction inlumen diameter at the point of the stenosis.Other methods have been used including esti-mates of the cross sectional area stenosis orsimply measurement of the residual lumendiameter. We have relied on experienced radi-ologists reporting their visual impression ofthe degree of stenosis present, as we believethat this is the method most commonly usedin routine clinical practice. The between andwithin observer variation in reporting withthis method may be expected to be greaterthan when caliper measurements are used.Caliper readings were also taken based onmeasurements of the diameter of the residuallumen and of the distal common carotidartery as described. The common carotidartery diameter was used as denominatorbecause the distal internal carotid artery was

1468



j.com/

J Neurol N


ecember 1994. D

ownloaded from



Figure 2 Differencesbetween ultrasound (US)and digital subtractionangiography (DSA)measurements plottedforeach vessel against the %stenosis as measured byDSA. For perfectagreement points would alllie along the zero %difference line.

100 _

80 ------- Mean difference

60 H

En0

0)C]

0)00r-

O- -4

40

-C

-ic

00 00 ~~~~~~~~2

200 0 0 0

43_

22_0 2 5

20 0

z II0 10 20 30 40 50 60 70 80 90 100

% Stenosis by DSA

often not included in the three dimensionalmagnetic resonance angiogram. These resultsare not therefore percentage reductions inlumen diameter, but are related, assuming aconstant relation between the diameter of thecommon carotid artery and the normal proxi-mal internal carotid artery. There is evidenceto suggest a closer relation between the com-mon carotid artery and the carotid bulb thanbetween the distal internal carotid artery andcarotid bulb."' Regardless of the relation tothe true percentage reduction in diameter, thecaliper results enable comparison of magneticresonance and digital subtraction angiogra-phy, because they are objective measurementsof the same vessel diameters as displayed byeach technique.

STATISTICAL ANALYSISStudies that compare one measurement tech-

nique with another are often analysed inap-propriately, notably by using correlationcoefficients.'2 Previous studies comparingmagnetic resonance angiography with conven-tional angiography have often quoted highcorrelation coefficients as an indication ofgood agreement. 13-16 The correlation coeffi-cient measures the strength of a relationbetween two variables, not the agreementbetween them, and it is possible to producehigh correlation coefficients for data that seemto be in poor agreement.'2 Problems also ariseif the data are divided into several groups-forexample, mild (0-29%), moderate (30-69%),severe (70-99%), and occluded (100%) cate-gories, with the frequency of assignments tothe same category by each method used as ameasure of agreement. Results by this tech-nique can be quoted simply as the proportionof assignments in agreement or as the Kappa

Figure 3 Differencesbetween magneticresonance angiography(MRA) and digitalsubtraction angiography(DSA) measurementsplotted, for each vessel,against the % stenosis asmeasured by DSA. Forperfect agreement pointswould all lie along the zero% difference line.

100 _

80 -------Mean difference

60 H

<4U)0

< 21

00)c

a. -2

o-4

-61

-81

-101

10 0

0 2 S 0 0 3* *00 3 2

!033_2 2 2 2 : 20~~~

!0 02.0~~~~~~~~~~

1011040l 60 7lll0 10 20 30 40 50 60 70 80 90 100

% Stenosis by DSA

1469

--I

IF



j.com/

J Neurol N


ecember 1994. D

ownloaded from



Figure 4 Differencesbetween caliper magneticresonance angiography(MRA) and caliper digitalsubtraction angiography(DSA) measurementsplotted, for each vessel,against the % stenosis asmeasured by the DSAcaliper results. For perfectagreement points would allbe along the zero %difference line.

100 _

80

LI)a

C.)cn)

a

0

C.)

C.)CL

CL)ci)ca)a)

r-o-0

-Mean difference

60 -

40 _

200

& 0* * . 0* ,2..- . o 0 - g 3*w3

°.8---------------; --- -- -. -w ---o A

-20 _

-40

-60

-80

-100-20 -10 0 10 20 30 40 50 60 70 80 90 100

% Stenosis by caliper DSA

statistic, which corrects for chance agree-ment.17 The results obtained by these tech-niques will vary according to which arbitrarycut off values are chosen and also according tothe number of categories chosen.'8 The valueof a statistic such as Kappa in assessing agree-ment, particularly of continuous data, istherefore limited. The method described byBland and Altman and Altman and Bland,which simply takes the differences betweenthe two tests and calculates the mean differ-ence and the SD of the differences is moreappropriate.'2 19 The mean difference is anestimate of the average bias of one methodrelative to another and would be significantlydifferent from zero in the case of one methodsystematically recording higher or lower val-ues. The SD is a measure of the error betweenthe methods'9; wide differences betweenmethods and an associated large SD indicates

poor agreement. The level of agreement andany relation to the degree of stenosis can beshown by plotting the differences against themeans for each pair of test results.'2 '9 Pooragreement would result in a wide scatter ofpoints about the mean difference, whereasany relation between agreement and percent-age stenosis would result in a variation in thescatter of points for different values of meanstenosis (figs 2-5). The 95% limits of agree-ment, the mean difference +2SD, has beensuggested as the most appropriate way toanalyse method comparison data.'2 As will beseen, when comparing the percentage stenosismeasured by different techniques, agreementvaries according to the degree of stenosisbeing measured. This is in part because whenresults are expressed as percentage stenosis,the maximum possible disagreement becomesprogressively less at both extremes of mean

Figure 5 Differencesbetween two radiologists(rad 1, rad 2) reportingthe same digital subtractionangiogram (DSA) filmsplottedfor each vesselagainst the % stenosis byDSA as used in theprevious plots.

100 _

80 -------Mean difference

60 H

-N

co

-

C.)C01)

L-

0)

0-.

0

00 00 0 3 0

* * 0L 2 0 3.02 0 0 0_0 5 0 03 3t 3~~~~---- -----

-- i

0 @ *2 * * 2 4 , , 2

2 2 0 00 0 ~~~~0 0

-20 2 0 0 v

-40

-60

-80

-1000 10 20 30 40 50 60 70 80 90 100

40 _

204

% Stenosis by DSA

1470



j.com/

J Neurol N


ecember 1994. D

ownloaded from



Table 1 Comparison of results by magnetic resonanceangiography with those by digital subtraction angiography

Digital subtraction angiography (%)Magnetic resonanceangiography (%) 0-29 30-69 70-99 100

0-29 32 11 0 030-69 7 13 7 070-99 0 3 48 3100 0 0 1 12

Values are absolute numbers.

stenosis, so that for a mean result of 0% or100%, there can be no difference between thetwo tests. By plotting the differences betweenmethods against the percentage stenosis asmeasured by digital subtraction angiography,this effect is eliminated. This approach is notnormally recommended because of an artefac-tual relation between the difference and eitherof the individual test results.'2 The relationbetween agreement and stenosis as measuredby digital subtraction angiography, however,is extremely important, as our current criteriafor management decisions are based on theresults ofECST and NASCET, and thereforeon x ray angiography. We have calculated themean difference (SD), which gives an indica-tion of the agreement on average, across thefull range of stenosis values. The closer themean difference to zero and the smaller theSD of the differences, the better the agree-ment. An idea of the specific agreement pre-sent at different degrees of stenosis can begained from inspection of the plots of differ-ences against percentage stenosis measured bydigital subtraction angiography.As well as measuring the agreement

between different methods of measurementwe also wish to describe the ability of the testto diagnose the actual degree of stenosis. Theaccepted "gold standard" in this context isintra-arterial x ray angiography although thistechnique has its own limitations. Any newmethod must be calibrated against the resultsof x ray angiography, especially as the ECSTand NASCET studies are based on resultsfrom the technique. Therefore, the sensitivity,specificity, positive predictive, and negativepredictive values for ultrasound and magneticresonance angiography have been calculatedat three clinically important cut off values-namely, 30%, 70%, and 100% stenosis-withthe assumption that the results of digital sub-traction angiography represent the true situa-tion. The relation between the caliper index,based on residual lumen and common carotidartery measurements, and the visual impres-sion of stenosis was assessed by linear regres-sion and Pearson's product moment correlation.

Table 2 Comparison of results by ultrasound with thoseby digital subtraction angiography

Digital subtraction angiography (%)

Ultrasound (%) 0-29 30-69 70-99 100

0-29 32 7 0 030-69 7 15 5 070-99 0 5 50 1100 0 0 1 14

Values are absolute numbers.

Table 3 Mean differences (SD) for digital subtractionangiography (DSA), magnetic resonance angiography(MRA), and ultrasound (US)

Mean differenceComparison % (95% CI) SD (%)

MRA-DSA -0-7 (-3l1 to 17) 14US-DSA 3-1 (0 6 to 5 6) 15MRA-US -3 8 (-6 3 to -1-2) 15Caliper (MRA-DSA) 0 9 (-0 3 to 2-2) 7

ResultsDuring the period of the study 83 patientswere referred for angiography. Of these, 79(95%) agreed to take part in the study. In twopatients (3%) it was not possible to selectivelycatheterise the common carotid arteries andfive patients (6%) were unable to completemagnetic resonance angiography. The reasonswere claustrophobia in two patients, develop-ment of chest pain in one patient withischaemic heart disease, severe back pain in apatient with chronic back pain, and onepatient being too large to fit in the tunnel ofthe scanner. One digital subtractionangiogram and one magnetic resonanceangiogram were considered uninterpretableand were excluded from the analysis. This left70 patients and a total of 137 vessels exam-ined by all three techniques. There were 49men and 21 women with a mean age of 62(range 37-76) years. Presenting symptomswere of recent, non-disabling stroke or tran-sient ischaemic attack in the carotid artery ter-ritory in all cases. There were 26 patients withstroke (37%), 36 with transient ischaemicattack (51 %) and eight (11 %) with bothstroke and transient ischaemic attack. Themean interval between ultrasound and digitalsubtraction angiography was 1-2 (range 0-8)days with 96% performed within two days.The mean interval between magnetic reso-nance angiography and digital subtractionangiography was 1-3 (range 0-8) days with92% performed within two days of each other.A total of 69 of 70 patients and 135 of 137vessels (99%) underwent two dimensionaland 55 of 70 patients and 107 of 137 vessels(78%) three dimensional magnetic resonanceangiography.

Tables 1 and 2 show the results for thecomparisons between magnetic resonanceangiography and ultrasound with digitalsubtraction angiography, after grouping intomild, moderate, severe, and occluded cate-gories.The plots of the differences against the

Table 4 Between and within observer variation fordigital subtraction angiography (DSA) and magneticresonance angiography (MRA)

Mean difference% (95% CI) SD (%)

Between observer variationDSA 2-9 (10 to 4 7) 11-2MRA 11 (-07to30) 10-8

Within observer variationDSA 2-4 (0 9 to 4-0) 9 0MRA 4 0 (2-3 to 5 7) 10-3Caliper DSA 0-8 (0 1 to 1-4) 3-9Caliper MRA 0-5 (-0-6 to 1-5) 5-7

1471



j.com/

J Neurol N


ecember 1994. D

ownloaded from



Table S Sensitivity, specificity, and positive and negative predictive values forultrasound (US) and magnetic resonance angiography (MRA) at different cut offpoints,assuming the results of digital subtraction angiography (DSA) to represent the truesituation

+ ve predicted -ye predictedSensitivity Specificity value value

Comparison (f%) (%) (%) (%)

For occlusionMRA 80 99 92 98US 93 99 93 99MRA and US* 92 100 100 99

For 70-100% stenosisMRA 90 95 96 90US 93 92 93 92MRA andUS* 94 95 95 94

For 30-100% stenosisMRA 89 82 93 74US 93 82 93 82MRA and US* 93 96 99 81

For 70-99% stenosisMRA 86 93 89 90US 89 93 89 93MRA and US* 92 95 92 95

*When in agreement (see text).

percentage stenosis as measured by digitalsubtraction angiography are shown for thecomparison between magnetic resonance

angiography and digital subtraction angiogra-phy, for ultrasound and digital subtractionangiography, and also for the caliper measure-ments (figs 2-4). Similar plots have been gen-

erated for between and within observervariation in reporting digital subtraction andmagnetic resonance angiograms. The differ-ences between two radiologists reporting thesame digital subtraction angiograms, plottedagainst the stenosis by digital subtractionangiography used in the other plots, is alsoshown (fig 5). Tables 3 and 4 show the meandifferences (SD) for these comparisons.

There is a good correlation between caliperresults for digital subtraction angiograms(DSA) and percentage stenosis measured byvisual impression of angiograms (caliper DSA= 0-71 DSA + 23, correlation coefficient, r =

0-91 (95% confidence interval 0-88 to 0 94)).Table 5 shows the sensitivity and specificity

results for ultrasound and magnetic resonance

angiography at three cut off values-namely,100%, 70%, and 30%-assuming the resultsof digital subtraction angiography to representthe true situation. Results are also shown forthose vessels in which ultrasound andmagnetic resonance angiography were inagreement concerning each specific cut off.Ultrasound and magnetic resonance were inagreement in 133 of 137 vessels (97%) forcomplete occlusion, 131 of 137 vessels (96%)for >,70% stenoses, and 119 of 137 vessels(87%) for > 30% stenoses. Similar resultshave been calculated for distinguishing opera-ble (70-99%) stenoses from non-operable(0-69% and 100%) stenoses according to thefindings, to date, of ECST and NASCET.Ultrasound and magnetic resonance agreed in127 of 137 vessels (93%) for this classifica-tion.

DiscussionThe results show good agreement between allthree techniques when assessing severe

carotid artery stenoses. There are some size-

able differences in the assessment of degreesof stenosis for some patients with mild tomoderate stenosis, in part related to interob-server variation in reporting. Based on digitalsubtraction angiography as the gold standard,non-invasive imaging will reliably identifythose patients with tight (70-99%) stenosis.The mean differences between the different

techniques were close to zero, indicating thatall three methods agreed closely on average,although sizeable differences occurredbetween patients (figs 2-4). When comparingdigital subtraction angiography and magneticresonance angiography the 95% confidenceinterval for the mean difference included zero,indicating no statistically significant bias. Thiswas not the case when comparing ultrasoundwith either digital subtraction angiography ormagnetic resonance angiography, indicatingthat ultrasound was recording slightly tighterdegrees of stenosis, on average, than the othertwo modalities. The SDs of the differencesbetween each pair of modalities, an estimateof the error between methods, were similar,indicating similar levels of agreement betweeneach pair of modalities. When comparing theresults measured by caliper the mean differ-ence was again close to zero whereas the SDof the differences was considerably less. Thusmeasuring the same diameters of vessels fromthe same viewing angle with calipers gavebetter agreement than recording the visualimpression of the percentage stenosis. Thecaliper measurements, based on the diameterof the common carotid artery some distanceaway from the point of maximum stenosis,did not actually measure the percentagestenosis, although there was a strong correla-tion between the two. The results of betweenand within observer comparisons show similarlevels of variation in the radiologists' reportingof digital subtraction and magnetic resonanceangiograms. The caliper measurements againshow less variation, although an importantpart of this analysis, namely between observervariation in caliper measurements, has notbeen performed. These results suggest that aconsiderable component to the disagreementbetween different methods is due to betweenand within observer variation in reporting.The application of quantitative computertechniques to aid lumen border definition,developed particularly for the analysis of coro-

20nary arteriograms, 0 could reduce this prob-lem in carotid angiography.The sensitivity, specificity, and positive and

negative predictive values calculated describethe ability of magnetic resonance angiographyor ultrasound to diagnose the percentagestenosis as measured by digital subtractionangiography, which we consider to representthe "true patient status". In fact, it is mostunlikely that digital subtraction angiographyperfectly describes the true situation in allcases, particularly given the variation inreporting that this and other studies haveshown."-24 The criteria currently shown toidentify which patients with cerebrovasculardisease should proceed to carotid endarterec-tomy are, however, based on the percentage

1472



j.com/

J Neurol N


ecember 1994. D

ownloaded from


1473Comparison of magnetic resonance angiography, duplex ultrasound, and digital subtraction angiography

Figure 6 (A) Intra-arterial digital subtractionangiogram showing tightstenosis ofproximalinternal carotid artery. (B)Two dimensional magneticresonance angiogram with"signal gap" in the regionofstenosis. (C) Threedimensional magneticresonance angiogram of thesame vessel showedsuperior signalfrom regionof stenosis.



j.com/

J Neurol N


ecember 1994. D

ownloaded from



Figure 7 (A) Intra-arterial digital subtractionangiogram reported asshowing complete occlusionof internal carotid artery(arrow). (B) Twodimensional magneticresonance angiogram of thesame vessel showing tightstenosis, with presence ofsignal gap, andflowwithin the distal internalcarotid artery (arrow).CCA = common carotidartery; ECA = externalcarotid artery;ICA = internal carotidartery.

A

Il-" A

CCA

carotid stenosis, as measured by conventionalx ray/digital subtraction angiography. There-fore, if we wish to manage patients appropri-ately by an alternative diagnostic test, we mustmeasure the performance of this test againstconventional angiography in the first instance.Because the diagnosis by conventional angio-graphy is not perfect, the sensitivity and speci-ficity values obtained for magnetic resonanceangiography will vary according to the preva-lence of the abnormality in the patients stud-ied.25 Strictly, these results should only beapplied to patients with the same prevalenceof disease as in this study. The resultsobtained show consistently high values ofsensitivity and specificity for both magneticresonance angiography and duplex ultrasoundcompared with digital subtraction angio-graphy.The specificity for both ultrasound and

magnetic resonance angiography in diagnos-ing occlusion was high at 99%. Each methoddiagnosed one vessel as occluded whereas thedigital subtraction angiogram showed a tightstenosis. In the case classified as occlusion bymagnetic resonance angiography, digital sub-traction angiography and ultrasound showed98% and 95% stenosis respectively. The mag-netic resonance angiogram was reviewed andthe misclassification found to be due to failurein distinguishing the distal internal carotidartery from branches of the external carotidartery. A characteristic feature of magneticresonance angiography is the appearance ofan apparent gap in the signal from vessels witha tight stenosis (fig 6). This signal gap corre-

sponds to a region of complex, non-laminarblood flow that exists for a short distanceimmediately beyond a tight stenosis and fromwhich little signal can be generated with theusual magnetic resonance angiographysequences. Further along the course of thevessel, as uniform blood flow reconstitutes,the signal returns, giving the appearance of agap in the vessel. Usually it is straightforwardto identify the signal from blood flow in themore distal internal carotid artery and to clas-sify the vessel as tightly stenosed. If the fulllength of the artery is not included in the scan,it is possible to miss signal from the distalinternal carotid artery, or to confuse signalfrom it as being from a branch of the externalcarotid artery, and misclassify the vessel asoccluded. On reviewing the individual axialslices from the two dimensional magnetic res-onance angiogram of the particular case, itwas evident that there was flow within the dis-tal internal carotid artery, indicating a tightstenosis. We now routinely review individualtwo dimensional slices through the skull basebefore diagnosing occlusion, looking for evi-dence of flow within the distal internal carotidartery. With this method, no cases of tightstenosis by digital subtraction angiographyhave been classified as occlusions by magneticresonance angiography. Classification of atight stenosis as an occlusion, or "pseudo-occlusion", has been a recognised problemwith duplex ultrasound,2627 which can beminimised by supplementing the duplexexamination with continuous wave probeexamination, and by the use of colour duplex

1474

-.ft...

.P.:-P.



j.com/

J Neurol N


ecember 1994. D

ownloaded from



sonography. Even with these techniques,some tight stenoses will still be diagnosed asocclusions by ultrasound.2829 The one vesselwith ultrasound pseudo-occlusion in thisstudy was recorded as 98% and 95% stenosedrespectively by digital subtraction angiogra-phy and magnetic resonance angiography.The sensitivity of magnetic resonance

angiography in diagnosing occlusion wouldseem to be low at 80%, with three vessels clas-sified as tightly stenosed by magnetic reso-nance angiography but as occluded by digitalsubtraction angiography. On reviewing thesecases it was considered that in two instancesthe magnetic resonance angiogram was cor-rect and that the digital subtractionangiogram had been misreported in one case,and had missed the tight stenosis in the other(fig 7). In this case the magnetic resonanceangiogram was straightforward to interpretand ultrasound examination, both before andafter digital subtraction angiography, clearlydocumented a tight stenosis. In the third case,a vessel was reported as tightly stenosed bymagnetic resonance angiography andoccluded by both digital subtraction angiogra-phy and duplex ultrasound. At the end of thestudy the magnetic resonance angiogram wasreviewed, still blinded to the results of otherinvestigations, and was reported as occluded.It is not clear why this vessel was originallymisclassified. Overall, particularly when indi-vidual axial two dimensional slices throughthe base of the skull are consulted, we believethat magnetic resonance angiography is highlyreliable in detecting occlusion.

For diagnosing stenoses of 70% or greater,both ultrasound and magnetic resonanceangiography perform well. Whatever themethod chosen, agreement between the twotests seemed to be better for stenoses > 50%than for stenoses <50%. This is shown clearlyin the plots of the differences against stenosisas measured by digital subtraction angiogra-phy (figs 2, 3), with small differences presentin the region of tight stenoses and consider-able differences for mild and moderatestenoses. A similar effect was also seen in theassessment of between and within observervariation for both magnetic resonance anddigital subtraction angiography (fig 5). In gen-eral there is more observer variation, andtherefore less agreement, when reporting mildstenoses, a finding noted in other studies.2' 23Consequently, both ultrasound and magneticresonance angiography are less sensitive andspecific in diagnosing mild stenosis than digitalsubtraction angiography. At these levels ofstenosis, ultrasound seems closer to theresults of conventional angiography.When only those cases in which the results

for ultrasound and magnetic resonanceangiography were in agreement are consid-ered, the comparison with digital subtractionis even closer. It has been suggested that itmay be possible to make decisions concerningprophylactic carotid endarterectomy with onlythe results from ultrasound and magneticresonance angiography, if these are in agree-ment.303' As the sensitivity and specificity of

this combination, measured against conven-tional angiography, is less than 100%, adoptingthis approach would result in some patientsbeing referred for operation, and some beingtreated medically, on the basis of non-invasivetests, when they would have been advised theopposite treatment had conventional angio-graphy been performed. For example, if it isassumed that only vessels classified as having70% to 99% stenosis should be consideredsuitable for prophylactic carotid endarterec-tomy then, from these results, magnetic reso-nance angiography classified 14 vessels (10%)differently from digital subtraction angiogra-phy (eight as inoperable and six as operable).Ultrasound similarly misclassified 12 vessels(9%), six as inoperable and six as operable.Considering just the 127 vessels (93%) inwhich ultrasound and magnetic resonanceangiography concurred regarding the decisionto operate, eight (6%) were classified differ-ently by digital subtraction angiography (fouras operable and four as inoperable). It is notnecessarily a requirement for non-invasiveimaging to be perfectly sensitive and specific,because a small loss of accuracy would beoffset by the absence of any appreciable riskin comparison with invasive angiography.Furthermore, given the considerable betweenand within observer variation in the reportingof angiograms that has been documented, it isclear that some patients are already beingmanaged inappropriately on the basis of con-ventional angiography. With the same criteriafor identifying vessels suitable for operation,there were disagreements over 13 vessels (9%)concerning suitability for operation when twodifferent radiologists were interpreting thedigital subtraction films, and eight disagree-ments (6% of vessels) when the same radiolo-gist reported the same digital subtraction filmson two different occasions. Thus when identi-fying vessels with 70% to 99% stenoses, thenumber of disagreements between magneticresonance angiography and digital subtractionangiography (14 vessels) was similar to thenumber produced by the between observervariation in reporting the digital subtractionangiograms (13 vessels).

Other investigators have reported goodresults with magnetic resonance, althoughstudies to date have been carried out on dif-ferent patient populations, with differenttechniques for conventional angiography,magnetic resonance angiography, and ultra-sound examination. The results from most ofthese studies are based on few patients (lessthan 50 bifurcations).'2'8 Others, due to therapid pace of technological development, arebased on magnetic resonance sequences thathave since been superseded,39 40 and somehave inappropriately used correlation to mea-sure agreement."3 14 Laster et al, in a largeseries, found excellent agreement betweenselective contrast angiography and twodimensional time of flight magnetic resonanceangiography, when comparing area stenosisby contrast angiography with visual appear-ance of magnetic resonance angiograms.4" Asdiscussed previously, both ECST and

1475



j.com/

J Neurol N


ecember 1994. D

ownloaded from



NASCET used estimates of percentage diam-eter reduction to assess stenosis. Huston et al,reported 100% sensitivity and 67% specificityfor two dimensional time of flight magneticresonance angiography in diagnosing stenosesof 50% or greater when compared with con-ventional angiography.4' By contrast with ourresults, they found that the degree of stenosiswas consistently overestimated by magneticresonance angiography.

Wherever possible, the results of both thetwo dimensional and the three dimensionalmagnetic resonance angiograms were avail-able to the radiologists. It is not possibletherefore, to compare separately the results ofthe two dimensional and three dimensionaltechniques with conventional angiography. Ingeneral we found the two techniques to becomplementary. The two dimensionalangiograms covered a wide region around thebifurcation from which it was easy to distin-guish the internal and external carotid arter-ies. This was not always the case whenviewing the three dimensional angiograms,which covered a smaller region. Interpretationof the three dimensional magnetic resonanceangiogram in this situation required cross ref-erence with the two dimensional magneticresonance angiogram for correct identificationof the internal carotid artery. The contrastbetween flowing blood and stationary tissue inthe two dimensional scan was generally high,even in cases of slow flow, a situation in whichvessels were poorly visualised by the threedimensional technique. The three dimen-sional technique provided higher resolution ofthe region of stenosis and gave higher signalfrom the complex flow beyond a tight stenosisthan the two dimensional technique.

There are some problems with magneticresonance angiography. Some patients simplycannot be scanned, such as those with pace-makers or severe claustrophobia. In this studyfive patients (6%) could not complete scans,two (3%) because of claustrophobia, althoughthis may have been a factor in five patientsdeclining to take part in the study.Occasionally the results obtained are uninter-pretable, most commonly because of excessivepatient motion. It is possible to place theinaging slices or volume in such a position asto miss the carotid artery bifurcation. In ourexperience this is usually due to poor visuali-sation of the bifurcation on sagittal localisingimages. We now use a series of axial twodimensional magnetic resonance angiogramslices covering the region from shoulder toskull base as localising images, from which itis relatively simple to locate the bifurcationand accurately position the full angiogramsequences. We do not routinely obtain viewsof the origin of the vessels from the aorticarch, and although such scans can beobtained, information concerning the accu-racy of magnetic resonance angiography inthis area is currently lacking. We routinelyobtain views of the carotid siphon, althoughimaging in this area poses technical difficultiesfor magnetic resonance angiography and dataconcerning accuracy in this region is again

lacking. Although there is some debate sur-rounding this particular problem,4344 in ourexperience relevant siphon stenosis is not acommon occurrence and rarely affects thedecision to operate; a view for which there isincreasing evidence.4546 Artifacts peculiar tomagnetic resonance angiography are encoun-tered, such as the signal gap occurring beyonda tight stenosis. As mentioned previously wehave found this useful as it indicates the pres-ence of a tight stenosis and others have evensuggested that this appearance could be usedto identify patients for endarterectomy with-out the need for any other preoperative stud-ies.47 In this study there were 42 vessels witha signal gap present on two dimensionalmagnetic resonance angiography. All of thesevessels were assigned to the severe stenosiscategory (70-99%) by digital subtractionangiography. Apparent signal gaps can occurwith the two dimensional magnetic resonanceangiogram techniques utilising a travellingsaturation slice. This most commonly occurswhen there is a loop in the artery, and creates apeculiar discontinuity in the apparent courseof the vessel. It results from arterial bloodtravelling in the opposite direction to normalin the neck-that is, towards the feet-and inso doing passing through a region of radio-frequency pulses positioned superiorly to theimaging slice for the purpose of preventingsignal generation from venous blood. Theloop can, however, be visualised quite clearlywith a three dimensional technique. A combi-nation of reversed flow and non-laminar flowcan occur in the normal carotid bulb givingrise to apparent flattening of the bulb onmagnetic resonance angiograms, more pro-nounced with the two dimensional technique.If this is not appreciated there will be a ten-dency to classify normal arteries as mildlystenosed. Apparent discontinuities can alsooccur if the patient moves appreciably duringthe acquisition of the two dimensionalangiogram. This is easy to identify and unlessthe slice being imaged at the time of motion isin a critical position, such as at the stenosis, itis not usually a serious problem. A recentlyformed thrombus can generate quite high sig-nals to the extent that it may show on theprocessed magnetic resonance angiogram.This could theoretically be mistaken for bloodflow, although the signal from thrombus has adifferent quality, which can be recognisedwith experience. The post-processing tech-nique of maximum intensity projection usedto convert the original axial sections into aprojection angiogram results in some loss ofthe lower intensity features of vessels, whichcan lead to apparent reduction in vessel diam-eter, overestimation of blood turbulence orstenosis, and poor visualisation of small ves-sels or vessels with slow flow.48An important finding from this and other

studies is that duplex ultrasound may be moreaccurate overall than magnetic resonanceangiography."" Several authors have sug-gested that decisions about operation could bemade on the basis of ultrasound assessmentalone.6 The benefit of including magnetic

1476



j.com/

J Neurol N


ecember 1994. D

ownloaded from



resonance angiography is that in some cases itwill be superior to ultrasound, as in demon-strating slow flow beyond a very tight stenosis.Also the full extent of the carotid system canbe displayed from origin to carotid siphonand intracranial vessels, with the bifurcationreliably seen in cases where ultrasound hasdifficulty, such as a behind the angle ofthe mandible. Finally, magnetic resonanceangiography provides a permanent morpho-logical record of the vessels, helpful and re-assuring to the surgeon in a way that valuessuch as doppler velocity and frequency shiftcannot be.The technique of examining the extracra-

nial vessels by ultrasound is highly dependenton skill and experience. It is therefore impor-tant that the ultrasound examination is per-formed by an experienced operator.

These results show that at high levels ofstenosis, non-invasive techniques of magneticresonance angiography and duplex ultrasoundcan categorise vessels in a similar way to digitalsubtraction angiography, particularly whenthe non-invasive techniques are themselves inagreement. As the proved criteria for recom-mending prophylactic carotid endarterectomyare based on identifying those patients with ahigh degree of stenosis, at present 70% ormore, reliable classification of patients can bemade on the basis of non-invasive imagingalone. Our current practice for patients withrecent non-disabling carotid territoryischaemic episodes, is to screen all patients byduplex ultrasound and those found to havehaemodynamically appreciable internalcarotid artery stenosis then proceed to mag-netic resonance angiography of the neck andhead. If both ultrasound and magnetic reso-nance examinations show a stenosis of 70% ormore then the patient is referred for prophy-lactic carotid endarterectomy. If the non-invasive tests are not in agreement or if anadequate examination is not possible then wewould recommend digital subtraction angio-graphy. This policy will reduce considerablythe number of patients undergoing invasiveangiography, particularly those with sympto-matic tight stenosis of the internal carotidartery. Such a policy has been assessed inother centres with good results.'0 31 In ourunit, the risk of digital subtraction angiogra-phy causing severe stroke in patients with sig-nificant carotid atherosclerosis is 2%, with arisk for carotid endarterectomy of 3%. Thecombination of ultrasound and magnetic reso-nance angiography before operation removesthe angiographic risk and represents animportant advance in the management ofcarotid stenosis.

We acknowledge the support of the Stroke Association, theDunhill Medical Trust, and the Mersey Regional HealthAuthority in this work.

1 European Carotid Surgery Trialists' Collaborative Group.MRC European Carotid Surgery Trial: interim resultsfor symptomatic patients with severe (70-99%) or withmild (0-29%) carotid stenosis. Lancet 1991;337:1235-43.

2 North American Symptomatic Carotid EndarterectomyTrial Collaborators. Beneficial effect of carotid

endarterectomy in symptomatic patients with high-gradecarotid stenosis. N EnglJ Med 199 1;325:445-53.

3 Wagner WH, Treiman RL, Cossman DV, Foran RF,Levin PM, Cohen JL. The diminishing role of diagnosticarteriography in carotid artery disease: duplex scanningas definitive preoperative study. Ann Vasc Surg 1991;5:105-10.

4 Farmilo RW, Scott DJA, Cole SEA, Jeans WD, HorrocksM. Role of duplex scanning in the selection of patientsfor carotid endarterectomy. BrJ Surg 1990;77:388-90.

5 Gertler JP, Cambria RP, Kistler JP, et al. Carotid surgerywithout arteriography: noninvasive selection of patients.Ann Vasc Surg 1991;5:253-6.

6 Ranaboldo C, Davies J, Chant A. Duplex scanning alonebefore carotid endarterectomy: a 5-year experience. EurJ7Vasc Surg 1991;5:415-9.

7 Murie JA, Morris PJ. Carotid endarterectomy in GreatBritain and Ireland. BrJ Surg 1986;73:867-70.

8 Hankey GJ, Warlow CP, Molyneux AJ. Complications ofcerebral angiography for patients with mild carotidterritory ischaemia being considered for carotidendarterectomy. J Neurol Neurosurg Psychiatry 1990;53:542-8.

9 Davies KN, Humphrey PR. Complications of cerebralangiography in patients with symptomatic carotidterritory ischaemia screened by carotid ultrasound. JNeurol Neurosurg Psychiatry 1993;56:967-72.

10 Humphrey P, Sandercock P, Slattery J. A simple methodto improve the accuracy of non-invasive ultrasound inselecting TIA patients for cerebral angiography. J NeurolNeurosurg Psychiatry 1990;53:966-71.

11 Williams MA, Nicolaides AN. Predicting the normaldimensions of the internal and external carotid arteriesfrom the diameter of the common carotid. Eur J7 VascSurg 1987;1:91-6.

12 Bland JM, Altman DG. Statistical methods for assessingagreement between two methods of clinical measure-ment. Lancet 1986;i:307-10.

13 Heiserman JE, Drayer BP, Fram EK, et al. Carotid arterystenosis: clinical efficacy of two-dimensional time-of-flight MR angiography. Radiology 1992;182:761-8.

14 Anderson CM, Saloner D, Lee RE, et al. Assessment ofcarotid artery stenosis by MR angiography: comparisonwith x-ray angiography and colour-coded Doppler ultra-sound. Am J Neuroradiol 1992;13:989-1003.

15 Wilkerson DK, Keller I, Mezrich R, et al. The comparativeevaluation of three-dimensional magnetic resonance forcarotid artery disease. J Vasc Surg 1991;14:803-1 1.

16 Wesbey GE, Bergan JJ, Moreland SI, et al.Cerebrovascular magnetic resonance angiography: acritical verification. J Vasc Surg 1992;16:619-32.

17 Cohen J. A coefficient of agreement for nominal scales.Educational and psychological measurement 1960;20:37-46.

18 Maclure M, Willett WC. Misinterpretation and misuse ofthe Kappa statistic. AmJEpidemiol 1987;126:161-9.

19 Altman DG, Bland JM. Measurement in medicine: theanalysis of method comparison studies. The Statistician1983;32:307-17.

20 Brown BG, Bolson EL, Dodge HT. Quantitative com-puter techniques for analyzing coronary arteriograms.Prog Cardiovasc Dis 1986;28:403-18.

21 Croft RJ, Ellam LD, Harrison MJG. Accuracy of carotidangiography in the assessment of atheroma of the internalcarotid artery. Lancet 1980;i:997-1000.

22 Brown PM, Johnston KW. The difficulty of quantifyingthe severity of carotid stenosis. Surgery 1982;92:468-73.

23 Murie JA, McKay AJ. Radiographic assessment of theextracranial internal carotid artery. J Cardiovasc Surg1986;27: 154-7.

24 Chikos PM, Fisher LD, Hirsch JH, Harley JD, Thiele BL,Strandness DE. Observer variability in evaluatingextracranial carotid artery stenosis. Stroke 1983;14:885-92.

25 Altman DG. Practical statistics for medical research. London:Chapman and Hall, 1991.

26 Ricotta JJ, Bryan FA, Bond MG, et al. Multicentre valida-tion study of real-time (B-Mode) ultrasound, arterio-graphy, and pathological examination. J7 Vasc Surg1987;6:512-20.

27 Comerota AJ, Cranley JJ, Katz ML, et al. Real-time B-mode carotid imaging. A three-year multicentre experi-ence. Jf Vasc Surg 1984;1:84-95.

28 Mattos MA, Hodgson KJ, Ramsey DE, Barkmeier LD,Sumner DS. Identifying total carotid occlusion withcolour flow duplex scanning. Eur Jf Vasc Surg1992;6:204-10.

29 Erickson SJ, Mewissen MW, Foley WD, et al. Stenosis ofthe internal carotid artery: assessment using colourDoppler imaging compared with angiography. Am JRadiol 1989;152:1299-305.

30 Turnipseed WLD, Kennell TW, Turski PA, Acher CW,Hoch JR. Combined use of duplex imaging and mag-netic resonance angiography for evaluation of patientswith symptomatic ipsilateral high-grade carotid stenosis. JfVasc Surg 1993;17:832-40.

31 Polak JF, Kalina P, Donaldson MC, O'Leary DH,Whittemore AD, Mannick JA. Carotid endarterectomy:preoperative evaluation of candidates with combinedDoppler sonography and MR angiography. Radiology1 993;186:333-8.

32 Polak JF, Bajakian RL, O'Leary DH, Anderson MR,Donaldson MC, Jolesz FA. Detection of internal carotidartery stenosis: comparison of MR angiography, colour

1477



j.com/

J Neurol N


ecember 1994. D

ownloaded from



Doppler sonography, and arteriography. Radiology1992;182:35-40.

33 Wagle WA, Dumoulin CL, Souza SP, Cline HE. 3DFTMR angiography of carotid and basilar arteries. Am JfNeuroradiology 1989;10:911-9.

34 Masaryk TJ, Modic MT, Ruggieri PM, et al. Three-dimensional (volume) gradient-echo imaging of thecarotid bifurcation: preliminary clinical experience.Radiology 1989;171:801-6.

35 Riles TS, Eidelman EM, Litt AW, et al. Comparison ofmagnetic resonance angiography, conventional angio-graphy, and duplex scanning. Stroke 1992;23:341-6.

36 Mattle HP, Kent C, Edelman RR, Atkinson DJ, SkillmanJJ. Evaluation of the extracranial carotid arteries: cor-relation of magnetic resonance angiography, andconventional angiography. J Vasc Surg 1991;13:838-45.

37 Kido DK, Barsotti JB, Rice LZ, et al. Evaluation of thecarotid artery bifurcation: comparison of magneticresonance angiography and digital subtraction archaortography. Neuroradiology 1991;33:48-51.

38 Furuya Y, Isoda H, Hasegawa S, Takahashi M, KanekoM, Uemura K. Magnetic resonance angiography ofextracranial carotid and vertebral arteries, includingtheir origins: comparison with digital subtraction angio-graphy. Neuroradiology 1992;35:42-5.

39 Kido DK, Panzer RJ, Szumowski J, et al. Clinicalevaluation of stenosis of the carotid bifurcation withmagnetic resonance angiographic techniques. ArchNeurol 1991;48:484-9.

40 Litt AW, Eidelman EM, Pinto RS, et al. Diagnosis ofcarotid artery stenosis: comparison of 2DFT time-of-

flight MR angiography with contrast angiography in 50patients. Am J Neuroradiol 199 1;12: 149-54.

41 Laster Jr. RE, Acker JD, Halford III HH, Nauert TC.Assessment of MR angiography versus arteriography forevaluation of cervical carotid bifurcation disease. Am JNeuroradiol 1993;14:681-8.

42 Huston III J, Lewis BD, Wiebers DO, Meyer FB, RiedererSJ, Weaver AL. Carotid artery: prospective blindedcomparison of two-dimensional time-of-flight MRangiography with conventional angiography and duplexUS. Radiology 1993;186:339-44.

43 Masaryk TJ, Obuchowski NA. Noninvasive carotidimaging: caveat emptor. Radiology 1993;186:325-8.

44 Polak JF. Noninvasive carotid evaluation: carpe diem.Radiology 1993;186:329-31.

45 Schuler JJ, Flanigan DP, Lim LT, Keifer T, Williams LR,Behrend AJ. The effect of carotid siphon stenosis onstroke rate, death, and relief of symptoms followingelective carotid endarterectomy. Surgery 1982;92:1058-67.

46 Mattos MA, van Bemmelen PS, Hodgson KJ, BarkmeierLD, Ramsey DE, Sumner DS. The influence of carotidsiphon stenosis on short- and long-term outcome aftercarotid endarterectomy. J Vasc Surg 1993;17:902-1 1.

47 Anson JA, Heiserman JE, Drayer BP, Spetzler RF.Surgical decisions on the basis of magnetic resonanceangiography of the carotid ..rteries. Neurosurgery 1993;32:335-43.

48 Anderson CM, Saloner D, Tsuruda JS, Shapeero LG, LeeRE. Artifacts in maximum-intensity-projection displayofMR angiograms. Am _J Radiol 1990;154:623-9.

Essential tremorDescriptions of essential tremor in literature have com-mented more on titubation than movement of thelimbs. Thomas Mann was clearly aware of the heredi-tary nature of the condition. I have not encountered adescription recognising the disorder's alcohol respon-siveness though no doubt Mrs Gamp would have beendelighted to justify her dependence on purely medici-nal grounds. George Stephenson might have beentempted to attribute his problem to the frustrations ofdealing with Sir Astley Cooper. That eminent surgeonhad opposed the London to Birmingham railwayas being too threatening to his property atBerkhampsted.

Charles Dickens, 1843-4, Martin ChuzzlewitMrs Gamp concluded by drawing her shawl tightlyover herself with both hands, and, as usual, referred toMrs Harris for full corroboration of these particulars.She had that peculiar trembling of the head which, inladies of her excitable nature, may be taken as a sureindication of their breaking out again very shortly...

Elizabeth Gaskell, 1853, CranfordThis effort at concealment was the beginning of thetremulous motion of head and hands which I haveseen ever since in Miss Matty.

Samuel Smiles, 1857, The lives of George and RobertStephensonAlthough he had for some time been in delicate health,and his hand shook from nervous affection, heappeared to possess a sound constitution.

George Eliot, 1860, The mill on the FlossHe wears a hare-skin on his chest, and has a tremblingin his talk...

Arnold Bennett, 1911, Hilda Lessways"And on Saturday afternoon too, when everybody'sabroad!" Sarah Gailey added gloomily, with her invol-untary small movements of the head. . ."Oh! my poordear!" Sarah Gailey moaned feebly, her head bobbingwith its unconscious nervous movements.

James Joyce, 1916, A portrait of the artist as a youngmanHe struck the flags again and tittered while his head

trembled with a slight nervous movement.

Thomas Mann, 1924, The magic mountainIt was this posture also which helped to steady anincipient shaking of the head, gave him his look ofbeing sternly reined up, and caused him to support hischin on his neckcloth in the manner so congenial to lit-tle Hans Castorp's taste. . Young Hans Castorpnoticed that the ancestral tremor brought on by his ill-advised walk continued to trouble him-he found itrather an embarrassment when in the dining-roomalmost as a regular thing now, his head would beginshaking at table; he found this impossible to preventand hard to dissemble. He tried various devices to dis-guise the weakness, for he could not continually sup-port his chin on his collar; he would keep his head inaction, turning it to the right and left in conversation, orbear hard against the table with the left forearm whenhe carried a spoonful of soup to his mouth, and sup-port his head with his hand. G D PERKIN

Regional Neurosciences Centre,Charing Cross Hospital,London W6 8RF, UK

NEUROLOGY IN LITERATURE

1478



j.com/

J Neurol N


ecember 1994. D

ownloaded from


Comparison angiography, duplex angiography internal · digital subtraction angiography was-0-7...

Documents

Transcript of Comparison angiography, duplex angiography internal · digital subtraction angiography was-0-7...