N e u r o r a d i o l o g y / H e a d a n d N e c k I m a g i n g • R ev i ew Provenzale and Sarikaya MRI, MRA, and CTA in Diagnosis of Artery Dissection

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

Neuroradiology/Head and Neck Imaging Review

James M. Provenzale1,2 Basar Sarikaya1,3 Provenzale JM, Sarikaya B

Comparison of Test Performance Characteristics of MRI, MR Angiography, and CT Angiography in the Diagnosis of Carotid and Vertebral Artery Dissection: A Review of the Medical Literature OBJECTIVE. Initial assessment of patients with suspected dissection of the carotid or vertebral arteries typically is made by MRI, alone or in combination with MR angiography (MRA) or CT angiography (CTA). We reviewed the medical literature to determine, based on test performance characteristics such as sensitivity, specificity, positive predictive value, and negative predictive value, whether evidence could be found to support routine use of one imaging technique over the other for assessment of suspected dissection. CONCLUSION. Test characteristics for MR techniques such as MRI and MRA were relatively similar to those for CTA in diagnosis of carotid and vertebral artery dissection.

D

Keywords: carotid artery, CT angiography (CTA), dissection, MR angiography (MRA), MRI, vertebral artery DOI:10.2214/AJR.08.1688 Received August 16, 2008; accepted after revision March 20, 2009. 1 Department of Radiology, Duke University Medical Center, Box 3808, Durham, NC 27710. Address correspondence to J. M. Provenzale. 2 Department of Radiology, Emory University School of Medicine, Atlanta, GA. 3 Department of Radiology, Gaziosmanpasa University, Tokat, Turkey.

AJR 2009; 193:1167–1174 0361–803X/09/1934–1167 © American Roentgen Ray Society

AJR:193, October 2009

issection of the craniocervical arteries (i.e., the carotid and vertebral arteries), which was once considered an uncommon diagnosis, has become increasingly recognized as a cause of stroke in young and middleaged individuals. In part, this increased recognition is due to greater awareness that dissection is often spontaneous in nature, thus raising suspicion of this diagnosis in patients who have not sustained neck or head trauma [1]. In addition, increased use of noninvasive imaging studies has allowed larger numbers of patients to be screened who might not otherwise have undergone invasive digital subtraction angiography (DSA) (i.e., catheter angiography) for diagnosis. In some populations, the rate of dissection of the craniocervical arteries has been diagnosed at rates 3–10 times greater than that determined before the use of MRI, MR angiography (MRA), and, more recently CT angiography (CTA) studies both in patients with spontaneous dissections and in those who have experienced blunt neck trauma [2–4]. At this point, a number of studies comparing test performance characteristics of noninvasive studies against those of DSA have been published. The role of Doppler sonography in diagnosis of dissection is not reviewed here because it is less commonly used and because limitations have been noted by multiple authors [2, 5]. The purpose of this review is to analyze the medical literature pertaining to comparisons of MRI with MRA

and CTA to DSA. In addition, this review sets out to determine whether, based on comparisons with DSA, the medical literature favors either CTA or MR techniques (i.e., MRI alone or in conjunction with MRA) for evaluation of the patient with suspected dissection of the craniocervical arteries. It is hoped that, from such a comparison, the reader can ascertain the relative merits of each imaging technique and assess the limitations of both. This review is not intended as a meta-analysis of the medical literature, which would include stratification of articles according to efficacy criteria, such as diagnostic-thinking efficacy and patient-outcome efficacy. Such a meta-analysis would appropriately include statistical analysis appropriate to assess the power of the studies used in the review. Materials and Methods We reviewed the medical literature on use of MRI, MRA, and CTA for articles reporting the use of these techniques for detection of dissection of the carotid and vertebral arteries. A MEDLINE search was performed using the search terms “carotid,” “vertebral,” “craniocervical,” “artery,” “dissection,” “sensitivity,” “specificity,” “MRI,” “MR angiography,” “helical,” “MDCT,” and “CT angiography.” This search was most recently performed on February 12, 2009. Only articles written in the English language were considered for review. This search revealed 66 articles that met the criteria, which were then reviewed. In addition, a wider search of published articles on the topic of carotid artery and vertebral artery dis-

1167

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

Provenzale and Sarikaya section was performed, in which another 25 articles on these topics were reviewed for indications of discussion of test performance characteristics for MRI, MRA, and CTA for detection of arterial injury. Articles that did not compare test performance of MRI, MRA, or MDCT against DSA as a reference standard were excluded, leaving a total of 21 articles (Table 1) for analysis. Test per-

formance characteristics were recorded and compared across studies. For the 21 articles of interest, the imaging study, number of patients undergoing evaluation by MR techniques or CTA, percentage undergoing DSA, number of dissections, sensitivity, and specificity as well as (when available) the positive predictive value (PPV) and negative predictive value (NPV)

were recorded (Table 1). To consider the effect of study design on test performance characteristics, we also noted whether a study was prospective or retrospective in nature (Table 1). Details of the imaging techniques used in each study were also recorded (Tables 2 and 3). In most articles, the PPV and NPV were not provided, but on occasion they could be calculated from data provided. Solely MRI

TABLE 1:  Sensitivity and Specificity of CT Angiography (CTA) or MR Angiography (MRA) for Detection of Cervicocephalic Arterial Dissection in Articles in Which These Techniques Were Compared With Digital Subtraction (Catheter) Angiography (DSA)

Article Müllges et al. [6], 1992

Imaging Methoda MRI

No. of Patients Percentage Undergoing of Patients Study MRI, MRA, Dissection Undergoing No. of Affected Design or CTA Type (No.) DSA Dissections Artery (No.) P

5

S

100

5

Sensitivity (%)

Specificity (%)

PPV NPV (%) (%)

C (5)

—

—

100

—

Kitanaka et al. [7], 1994

MRI

R

5

NS

100

5

V (5)

—

—

100

—

Levy et al. [8], 1994

MRI, MRA

P

180

NS

100

24

C (19), V (5)

79b

99

—

—

Zuber et al. [9], 1994

MRI

R

15

NS

100

19

C (9), V (10)

100

—

—

—

Leclerc et al. [15], 1996

CTA (1)

I

16

NS

100

18

C (18)

100

85

—

—

Ozdoba et al. [10], 1996

MRI

R

31

NS

48

12

C (12)

—

—

80

—

Stringaris et al. [11]c, 1996

MRI, MRA

I

12

NS

100

12

C (12)

100

100

—

—

Yoshimoto and Wakai [12], 1997

MRI, MRA

R

11

S

100

13

V (13)

73

—

—

—

Mascalchi et al. [13], 1997

MRI, MRA

R

14

S (10), T (4)

86

12

V (12)

100

—

—

—

Auer et al. [14], 1998

MRI, MRA

R

19

S (7), T (12)

84

17

V (17)

95

29

—

—

Múnera et al. [16], 2000

CTA (1)

P

60

T

100

10

C (7), V (3)

90

100

100

98

Biffl et al. [25], 2002

MRI, MRA

P

16

T

100

4

NS

75

67

43

89

CTA (1)

P

46

100

22

Miller et al. [26], 2002

CTA (4)

P

143

100

47

C (17), V (30)

MRI, MRA

P

21

Chen et al. [17], 2004

CTA (4)

R

17

Bub et al. [18]d, 2005

CTA (1, 4, 8)

R P

Schneidereit et al. [19], 2006 CTA (8)

T

68

67

65

70

51

99

—

—

100

21

C (4), V (17)

50

98

—

—

S (15), T (2)

100

19

V (19)

100

98

95

100

32

T

100

17

C (12), V (5)

76

93

68

95

170

T

14

15

NS

—

—

65

—

Utter et al. [20]e, 2006

CTA (16)

R

372

T

30

7

C (4), V (3)

—

—

—

92

Eastman et al. [21], 2006

CTA (16)

P

162

T

90

46

C (20), V (26)

98

100

100

99

Malhotra et al. [22], 2007

CTA (16)

P

119

T

77

26

C (13), V (13)

74

86

65

90

Pugliese et al. [23], 2007

CTA (4)

R

15

S (12), T (3)

100

15

V (15)

100

95

94

100

Sliker et al. [24]f, 2008

CTA (16)

R

108

T

100

NS

NS

—

—

WB CTA: 69, 74g WB CTA: 82, 91 Neck: 64, 68 Neck: 94, 100

Note—PPV = positive predictive value, NPV = negative predictive value, P = prospective, R = retrospective, I = indeterminate, S = spontaneous dissection, NS = not specified, T = traumatic dissection, C = carotid artery, V = vertebral artery, WB = whole-body, dash indicates not applicable. aNumbers in parentheses indicate single-detector or 4-, 8-, or 16-MDCT scanner. b100% sensitivity for carotid dissection and 20% sensitivity for vertebral dissection. cMRI and MRA were compared with DSA but DSA did not serve as reference standard. DSA had a sensitivity of 91% and a specificity of 100%. d Values cited in this article represent combined test characteristics for three readers for both carotid artery dissection and vertebral artery dissection, although they are reported separately for individual readers and for type of artery in the article cited. eDSA was solely performed in patients with normal and equivocal CTA studies. Authors were primarily interested in negative predictive value; 271 patients with normal CTA were further evaluated, and 82 underwent DSA, which was positive in seven. Authors relied on reports for imaging findings, except retrospective review of false-negative case. fAuthors compared two different CTA protocols with DSA and their analysis was based on (arbitrarily) designated vessel segments and thus total number of dissections. The distributions could not be deduced. Sensitivities and specificities are given for vessel segments for each protocol. Authors relied on DSA reports to extract results. gValues are for cervical carotid artery and vertebral artery, respectively.

1168

AJR:193, October 2009

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

MRI, MRA, and CTA in Diagnosis of Artery Dissection or MRI in combination with MRA was compared with DSA in nine articles [6–14], solely CTA was compared with DSA in 10 articles [15–24], and both CTA and MRA were compared with DSA in two articles (with CTA generally being performed in a different group of patients than the group undergoing MRA in the same study) [25, 26].

Results The 21 articles that met the inclusion criteria were published over a 17-year interval. Three CTA studies were performed us-

ing single-detector scanners [15, 16, 25], and nine studies were performed using either solely MDCT scanners [17, 19–24, 26] or (in one case) predominantly MDCT scanners [18] (Tables 1 and 3). We did not include one article that reported four dissections but in which DSA was performed solely in two patients [27]. The number of patients undergoing either CTA or MR techniques in any one study varied between five and 372. The percentage of patients undergoing DSA in any one study varied between 14% and 100%,

with all patients undergoing DSA in 14 articles. As the table shows, in some articles, solely the sensitivity and specificity (and not the PPV and NPV) were provided. In one article, the reader could derive the PPV (but no other test performance characteristic) from data presented [19]. The articles listed in Table 1 vary considerably in the amount of information regarding age of dissections. In eight articles in which imaging was performed for assessment of neck injury, no specific information

TABLE 2:  Features of MR Scanners Used in Cited Articles

Article

MR Scanner Imaging Field Strength Technique (T)

MRI Sequences

MRA Contrast Material Administration

Technique of Unenhanced MRA

Technique (2D vs 3D)

Slice or Slab Thickness (mm)

Müllges et al. [6], MRI 1992

1.5

T1W, T2W, PDW

MRA not performed

MRA not performed

MRA not performed

5

Kitanaka et al. [7], 1994

MRI

0.5

T1W, T2W

MRA not performed

MRA not performed

MRA not performed

5–11

Levy et al. [8], 1994

MRI, MRA

Zuber et al. [9], 1994

MRI

Ozdoba et al. [10], 1996

MRI

1

MRI: 7; MRA: 1

Matrix Size and Field of View (FOV) Not specified Matrix, 192 × 256; FOV, 25 cm

Sagittal and axial Unenhanced SE T1W, axial SE T2W

FISP (additional 3D 3D TOF for head in eight patients)

0.5

Axial T1W and T2W

MRA not performed

MRA not performed

MRA not performed

1.5

Sagittal T1W; MRA not axial T1W, PDW, performed T2W; axial and coronal fat-saturated T1W

MRA not performed

MRA not performed

4-mm slice Not specified thickness with 1.5-mm gap and thin sections (3-mm slice thickness with 0.5-mm gap) for fat-saturated T1W sequences

Yoshimoto and MRI, MRA Not specified Wakai [12], 1997

Not specified

Not specified

Not specified

Not specified

Not specified

Mascalchi et al. [13], 1997

MRI, MRA

0.5

Sagittal T1W, axial PDW and T2W with additional sequences in majority of patients

Unenhanced

PC and TOF

PC: 2D (two patients), TOF: 2D (four patients) and 3D (one patient)

MRI: 5–7; PC: MRI: not specified; 20–40 (coronal PC: matrix, 192 × or sagittal); 2D 224; FOV, 20–25 TOF: 3; 3D TOF: 1 cm; 2D TOF: matrix, 160 × 192; FOV, 20–25 cm; 3D TOF: not specified

Auer et al. [14], 1998

MRI, MRA

1.5

T1W, PDW, and T2W in at least 2 orthogonal planes

Unenhanced

TOF (FISP) for head and 3D FLASH (coronal) for neck

3D (additional coronal 2D TOF in some patients)

MR: not specified; Not specified 3D TOF: 1; 2D TOF: not specified; FLASH: not specified

Biffl et al. [25]a, 2002

MRA

1. 5

Not performed

Unenhanced and contrastenhanced (3D FLAIR)

TOF

2D and 3D TOF, 3D FLAIR

Not specified

Not specified

Miller et al. [26]a, MRA 2002

0.2

Not performed

Unenhanced

TOF

2D

Not specified

Not specified

4–6

MRI: matrix, 230 × 256; FOV, 230 mm; MRA: matrix, 230 × 256; FOV, 230–250 mm Not specified

Not specified

Note—MRA = MR angiography, TW1 = T1-weighted imaging, TW2 = T2-weighted imaging, PDW = proton density–weighted imaging, SE = spin-echo, FISP = fast imaging with steady-state precession, TOF = time-of-flight MRA, PC = phase contrast MRA. aFor studies that implemented both MR techniques and CTA, solely MR parameters are given here.

AJR:193, October 2009

1169

Provenzale and Sarikaya TABLE 3:  Features of CT Scanners Used in Articles Cited

Collimation (mm)

Slice Thickness and Reconstruction Increment (mm)

Pitch

Leclerc et al. [15], Helical 1996

2

1

Not specified

3 mm/s

Supine, head tilted back, Not specified Not specified C2–C6; total coverage: 90 mm

Múnera et al. [16], Helical 2000

3

1

1.33

4 mm/s

C7 to base of mandible

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

Article

CT Equipmenta

Biffl et al. [25]b, 2002

Helical

Miller et al. [26]b, 2002

MDCT (4)

1

1

2:1

Chen et al. [17], 2004

MDCT (4)

1.25

0.8

Not specified

Bub et al. [18], 2005

Helical and MDCT (4, 8)

Not specified

1,1 interval

2

Position and Imaging Volume

Table Speed

Voltage (kVp)

120

Tube Current (mAs)

250

Not specified C3 to sella turcica

Not specified Not specified

Not specified Aortic arch to skull base

Not specified Not specified

7.5 mm per rotation, 9.4 mm/s

Aortic arch to circle of Willis

120

280

2.5 (MDCT protocol)

1–3 (varied 1.35:1 (MDCT depending on the protocol) CT equipment)

Not specified Aortic arch to circle of Willis

Not specified Not specified

Schneidereit et al. MDCT (8) [19], 2006

Not specified

1.25, 0.9 interval

1.35:1

13.5 mm/s/per Aortic arch to circle of rotation Willis

Not specified Not specified

Utter et al. [20], 2006

MDCT (16)

Not specified

2. 5

Not specified

Not specified Aortic arch to circle of Willis

Eastman et al. [21], 2006

MDCT (16)

Not specified

1.25, 0.5 interval

Not specified

Not specified Aortic arch to vertex

Not specified Not specified

Malhotra et al. [22], 2007

MDCT (16)

2

Not specified

Not specified Aortic arch to circle of Willis

Not specified Not specified

Pugliese et al. [23], 2007

MDCT (4)

1.25

1.5

Sliker et al. [24], 2008

MDCT (16)

0.75 Not specified 0.75

0.5

7.5 mm/per rotation

Not specified

140

120

380

230

Whole-body, Not specified Whole-body: arms Not specified Not specified 1.0; neck, 0.9 abducted, circle of Willis to lung bases or ischial tuberosities; neck: arms adducted, aortic arch to circle of Willis

aNumbers in parentheses indicate single-detector or 4-, 8-, or 16-MDCT scanner.

bFor studies that implemented both MR techniques and CTA, only CTA parameters are given here.

regarding age of dissection was provided [16, 18–22, 25, 26]. However, presumably imaging was performed within a few days of injury. Eight articles dealing with either spontaneous dissections or traumatic dissections reported findings on dissections that were thought to range between 3 days and 6 weeks old [7–9, 12, 13, 15, 17, 24]. Information regarding the age of dissections was not available in three articles [11, 14, 23]. This study assessed dissections in all segments of the carotid and vertebral arteries. Articles rarely stated locations of dissections along the length of the involved artery, so it is not possible to determine how many dissections were intracranial as opposed to extracranial. Nonetheless, based on images provided and the discussion within all articles, it appears that the overwhelming majority of dissections were in the extracranial segments.

1170

MR Techniques for Evaluation of Carotid Artery and Vertebral Artery Dissection For purposes of this review, all articles that, in any manner, compared MR techniques with DSA are discussed together. Thus, articles that solely compared MR techniques with DSA [6–14] will be discussed along with the MRI portions of articles that compared both CTA and MR techniques against DSA [25, 26]. In these 11 studies, a total of 329 patients were evaluated (with approximately 55% enrolled in one study [8]) and 144 dissections were reported. The reported sensitivity of MR techniques for detection of arterial dissection ranged from 50% to 100%. As Table 1 indicates, the specificity rate also varied considerably across studies, from 29% to 100%. In the two studies that compared both CTA and MR techniques with DSA, the sen-

sitivity and specificity for CTA and MR techniques were similar to one another (Table 1). One study reported sensitivity for CTA and MR techniques of approximately 50% and a specificity of approximately 98% [26]. In another study, both CTA and MR techniques had a sensitivity and specificity of approximately 70%; the PPVs for CTA and MR techniques were 65% and 43%, respectively, and the NPVs for CTA and MR techniques were 70% and 89%, respectively [25]. The largest series using MRI and MRA, which reported 24 dissections on DSA in 180 patients, showed good sensitivity and specificity rates of 83% and 99%, respectively [8]. Notably, in that study, sensitivity and specificity of MRI with MRA differed substantially for carotid artery dissection compared with vertebral artery dissection. For carotid dissections, MRI and MRA had a sensitivity of 95%

AJR:193, October 2009

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

MRI, MRA, and CTA in Diagnosis of Artery Dissection and a specificity of 99%. For vertebral artery dissection, MRI and MRA had a sensitivity of 20% and a specificity of 100%. In one small study of 12 patients with dissection, MRI and MRA were compared with DSA; however, in that study, DSA was not considered as a reference standard [11]. Instead, in cases in which the DSA study was interpreted as negative and the combination of MRI and MRA was considered positive, the MR studies were considered correct. DSA was determined to have a sensitivity of 91% and a specificity of 100%. Importantly, the PPV, which is the proportion of patients with dissections who were correctly diagnosed, was reported in only one article, in which it was found to be only 43% [25]. In that same study, NPV was reported to be 89% [25]. The remaining articles did not provide data that allowed these important values to be calculated. Sensitivity values for retrospective studies using MR techniques, which ranged from 95% to 100%, were higher than for prospective studies, which ranged from 50% to 79%. Specificity values for prospective studies ranged from 67% to 99%; the only specificity value available for a retrospective study was 29%. Because a PPV and an NPV were available for only one article involving MR techniques, comparison between retrospective studies and prospective studies was not possible for those indices. CTA for Evaluation of Carotid Artery and Vertebral Artery Dissection The number of studies for which CTA has been compared with DSA was found to be much larger than those using MRI and MRA, with substantially more patients reported in the CTA studies. A total of 1,277 patients underwent CTA in the combined studies, and 242 arterial dissections were found on DSA. In one article, the number of dissections could not be deduced [24]. The sensitivity values ranged between 51% and 100%, and the specificity rates varied between 67% and 100%. Notably, three of the 10 studies that reported sensitivity found a sensitivity rate of 100% [15, 17, 23] and five of the 10 studies that reported specificity indicated a specificity rate of at least 95% [16, 17, 21, 23, 26]. A PPV was reported in eight studies and ranged from 65% to 100% [16–19, 21–23, 25]. An NPV was reported in eight series and varied between 70% and 100% [16–18, 20–23, 25]. Among the CTA articles we reviewed, 25% were performed using single-detector CT scanners [15, 16, 25]. Comparing results obtained using MDCT scanners and those

AJR:193, October 2009

obtained using single-detector scanners, one finds that a moderate improvement in some (but not all) of the test characteristics has been reported with MDCT scanners. Sensitivities for retrospective CTA studies ranged from 64% to 100%, and those for prospective studies ranged from 51% to 98%. Specificities for retrospective CTA studies ranged from 82% to 95%, and those for prospective studies ranged from 67% to 100%. PPVs for retrospective CTA studies ranged from 68% to 94%, and those for prospective studies ranged from 65% to 100%. NPVs for retrospective CTA studies ranged from 92% to 100%, and those for prospective studies ranged from 70% to 98%. Thus, the range of values for any single test performance characteristic was similar for retrospective CTA studies and for prospective CTA studies. Discussion This review showed that the reported sensitivity and specificity of MR techniques and CTA for diagnosis of craniocervical arterial dissection are relatively similar. In this review, with one exception, only articles that compared MR techniques or CTA with DSA as a reference standard were considered. As some authors have pointed out, DSA is an imperfect reference standard because of interobserver variability and a number of entities that can mimic arterial dissection [23]. Nonetheless, the choice of solely articles in which DSA served as a comparison study for this review can be justified because, at least for the present, DSA is widely considered by the radiologic community as the reference standard for cross-sectional angiographic imaging studies. In the future, it is possible that a noninvasive imaging technique may serve as a reference standard, building on the previous small study suggesting that MR techniques are superior to DSA [11]. However, for that to be the case, a large study comparing techniques will be needed. This review also showed that the reported PPVs and NPVs for CTA are very good. Unfortunately, because PPV and NPV have rarely been reported for studies using MR techniques, a comparison with CTA with regard to those values cannot reasonably be performed. One limitation we found in attempting to compare studies using MRI techniques is that the age of the dissections in patients was often poorly outlined. This fact confounds an attempt to determine reasons for differing test performance characteristics between imaging tests for dissection. The age of the dissection is an important consideration because the de-

gree of bright signal on T1- and T2-weighted MR images, and thus one factor determining the conspicuity of dissections on MRI, is dependent on the age of the dissection. However, the age of dissection is difficult to determine in cases of spontaneous dissection. Often, one can only assume that the time of onset of dissection corresponds to the time of onset of symptoms and signs; however, this assumption may not be true in all cases. MRI and MRA for Evaluation of Carotid and Vertebral Artery Dissection The test performance characteristics reported in the studies using MR techniques varied substantially. Many possible explanations could account for this fact. Study design, experience of readers interpreting scans or DSA studies, available MR technology, postprocessing, and workstation capability are a few of the reasons that are likely important. One factor to be considered in evaluating variation between studies is whether solely a clearly abnormal study is considered a positive imaging study or whether equivocal studies, which in clinical practice would lead to further imaging, should also be considered. For instance, one investigator, noting that recognition and evaluation of equivocal findings is important for maximizing sensitivity, argues that equivocal findings should be considered as positive studies [18]. Such a practice would also conform to clinical practice in that equivocal studies are typically followed by a more definitive test. Because MRI and MRA were developed earlier than CTA techniques, comparisons of DSA against MRI and MRA were conducted earlier than CTA comparisons. No article indicating test performance characteristics for MRI and MRA published after 2002 was found, compared with eight CTA studies that provided information after 2002. Thus, newer techniques such as bolus-timing and timeresolved contrast-enhanced MRA are not included. Unfortunately, the test characteristics of MRA using more recent techniques have not been determined. Thus, information on trends in test performance of MRI and MRA over the past few years, after development of the aforementioned technologic advances, is lacking, whereas (as discussed later) such trends can be followed for CTA. CTA for Evaluation of Carotid and Vertebral Artery Dissection This review found that sensitivity of CTA in some studies in which it was compared

1171

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

Provenzale and Sarikaya with DSA is lower than in some CTA studies in which tests other than DSA (e.g., clinical examination or surgical findings) served as the sole reference standard (outlined later). A possible explanation for this discrepancy in sensitivity values is that DSA is more sensitive than these other tests for subtle dissections. Examples of studies in which the sensitivity and specificity of CTA were determined primarily on the basis of clinical follow-up include the following articles: In one study of 106 patients assessed with a 4-MDCT scanner in whom the final diagnosis was determined by various forms of imaging, surgical procedures, or clinical follow-up, the reported sensitivity for detection of vascular and aerodigestive tract injuries was 100% and specificity was 94% [28]. In another prospective study of 175 patients, the reference standard for positive CTA studies was catheter angiography or surgery, and for negative studies clinical follow-up served as the reference standard [16]. The reported sensitivity was 100% and specificity was 99%. Finally, in another study examining the use of CTA for detection of arterial injury in patients with blunt cervical trauma, MRA was the primary reference standard; among the 11 arterial injuries noted among 407 patients, sensitivity was recorded as 100% and specificity as 99% [2]. As with studies using MR techniques, marked variation was seen in test performance characteristics for CTA studies. Again, a number of possible explanations exist for these differences between studies, which have been outlined in the earlier discussion of variations in MRI and MRA test performance. As an example of the possible role of reader experience, in one study, all false-negative CTA examinations were recorded in the first half of the study, indicating the likelihood that test performance, in some cases, is substantially dependent on the level of experience of the individual interpreting the study [22]. Other potential reasons for variability in test performance characteristics, specifically for CTA studies, are advances in technology. In the past decade, considerable advances have been made in the design of helical CT scanners, e.g., development of MDCT capability on the order of 16 or more slices, improvements in contrast bolus–timing mechanisms, and developments in image processing programs and workstations. For instance, CTA data can be viewed as source images, maximum-intensity-projection images, volume-rendered (3D reconstruction) images,

1172

and multiplanar reformation images. Furthermore, much greater (and wider) experience has been gained in the last decade by radiologists in assessment of CTA imaging studies. Thus, it is natural to ask whether the performance characteristics for CTA have improved over time. For instance, in one study, investigators compared results of screening patients with blunt cervical trauma obtained using 16-MDCT helical CTA with those previously obtained using 4-MDCT helical CTA and found an approximately threefold increase in incidence of carotid or vertebral artery dissection using 16-MDCT CTA [29]. However, these differences, although suggestive of the importance of technologic advances in providing greater accuracy, could be explained by a number of other factors, e.g., differences in patient populations screened by 4-MDCT and 16-MDCT techniques or increased familiarity of radiologists with CTA images. Nonetheless, the comparison of test performance characteristics for single-detector CTA and MDCTA studies outlined in Tables 1 and 3 shows, in general, better test performance characteristics for MDCT scanners than for single-detector scanners. The specificity and NPV were generally higher for MDCT studies. These improved test characteristics are concordant with the reported improved test characteristics of high-resolution MDCTA for other purposes, e.g., for detection of carotid stenosis [30]. Although this review showed that the test characteristics of studies using MR techniques and CTA were relatively similar, some investigators have indicated a preference for CTA for diagnosis of dissection. For instance, some authors have pointed out that limitations found using MRI and MRA (e.g., obscuration of mural hematoma by the hyperintense signal of thrombus within an occluded artery and lack of hyperintense signal within mural hematoma on T1-weighted images in the first few days after onset of dissection) are overcome by CTA [31]. Another study (which did not compare CTA or MRA against DSA and hence is not included in our tables) found that CTA frequently depicted more findings indicative of dissection (e.g., more frequently found intimal flaps and pseudoaneurysms) than MRI and MRA [32]. It may be asked under what circumstances DSA should be performed for dissection, when noninvasive tests without patient risk are available. The first circumstance involves cases in which CTA, MR techniques, or both provide discrepant results. For instance, on

MR studies, when some pulse sequences suggest the presence of dissection and other sequences indicate absence of dissection, it may not be clear which sequences should be taken as representing truth. Depending on the level of comfort with CTA of the radiologist interpreting the imaging findings, DSA may be the most appropriate next study. Similarly, when discrepant or equivocal results are seen on CTA, the radiologist’s experience with MR techniques may favor DSA as the most suitable next study. The second circumstance involves acute clinical situations in which, if a dissection is found, an endovascular procedure (e.g., intraarterial stent placement) will likely be needed soon afterward. In such cases, in the interest of time, DSA could reasonably be deemed the most appropriate diagnostic study. Finally, it is, of course, indicated that, if DSA is already being performed on vessels other than the carotid or vertebral arteries and dissection of the craniocervical arteries is suspected, then those arteries are also included as part of the DSA procedure. Limitations This review was limited in its conclusions by the heterogeneous nature of the medical literature on the use of MR techniques and CTA in assessment of arterial dissection. We recognize the many limitations inherent in a review in which many different types of studies were pooled. Studies substantially differed from one another in degree of detail regarding such features as patient characteristics, ages of dissections, and cause of dissections. Furthermore, a great degree of variability in study design, methods of selection of patients, scanner capability, and imaging parameters and techniques was seen. For these reasons, conclusions regarding a comparison of MR techniques and CTA for diagnosis of arterial dissection of the cranio­ cervical arteries must be viewed with caution. In many instances, study features that could strongly influence test performance characteristics (e.g., potential sources of bias in studies) are not evident to the reader, making comparison of studies difficult. It is worth noting that many of the articles discussed in this review pertain to traumatic dissection rather than spontaneous dissection. This fact is particularly true for articles written from the perspective of trauma surgeons or emergency department physicians seeking to understand the use of imaging as a screening tool for traumatic carotid or verte-

AJR:193, October 2009

MRI, MRA, and CTA in Diagnosis of Artery Dissection

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

bral artery dissection. Thus, it is possible that the cumulative findings discussed here may not adequately reflect rates of detection for spontaneous dissections. Summary In conclusion, this review of the medical literature showed that, in general, the test characteristics of MRI combined with MRA were quite similar to those of CTA for diagnosis of dissection of the carotid and vertebral arteries. Clearly, both imaging techniques have unique advantages as well as pitfalls that serve to diminish sensitivity and specificity. However, the evidence obtained from this review does not indicate superiority of one technique as applied to a population of patients with carotid or vertebral artery dissection. Optimally, a formal prospective study with blinded readers and a well-defined patient population that might better compare imaging techniques would answer the questions we raised. However, it is unlikely that such a study will be performed in the future. Therefore, it is reasonable that individual users exercise their own preference based on specific clinical situations (e.g., ease of obtaining the imaging study, ability to couple the imaging study with other imaging tests so as to conserve time in managing patients) and degree of comfort with one particular technique. At our institution, CTA is routinely chosen for imaging the trauma patient because it allows multiple sites of injury (including those outside the head and neck) to be evaluated on a single scanner in close temporal order. However, that choice is based on consensus among the neuroradiologists at our institution rather than on evidence culled from the medical literature indicating a clear superiority of CTA for diagnosis of dissection. At our institution, patients with suspected arterial dissection that is not related to trauma are evaluated by either MRI combined with MR angiography or CTA in relatively equal numbers. On the basis of our review, it appears that individual readers, in conjunction with physicians ordering studies, may choose the study technique based on individual factors such as urgency of patient’s clinical condition in relation to scanner availability, coexisting diagnoses under consideration, and degree of radiologist experience with imaging techniques in this setting. References 1. Schievink WI. Spontaneous dissection of the carotid and vertebral arteries. N Engl J Med 2001;

AJR:193, October 2009

344:898–906 2. Mutze S, Rademacher G, Matthes G, Hosten N, Stengel D. Blunt cerebrovascular injury in patients with blunt multiple trauma: diagnostic accuracy of duplex Doppler US and early CT angiography. Radiology 2005; 237:884–892 3. Rogers FB, Baker EF, Osler TM, Shackford SR, Wald SL, Vieco P. Computed tomographic angiography as a screening modality for blunt cervical arterial injuries: preliminary results. J Trauma 1999; 46:380–385 4. Biffl WL, Egglin T, Benedetto B, Gibbs F, Cioffi WG. Sixteen-slice computed tomographic angiography is a reliable noninvasive screening test for clinically significant blunt cerebrovascular injuries. J Trauma 2006; 60:745–752 5. Berne JD, Norwood SH, McAuley CE, Villarreal DH. Helical computed tomographic angiography: an excellent screening test for blunt cerebrovascular injury. J Trauma 2004; 57:11–19 6. Müllges W, Ringelstein EB, Leibold M. Non-invasive diagnosis of internal carotid artery dissections. J Neurol Neurosurg Psychiatry 1992; 55: 98–104 7. Kitanaka C, Tanaka J, Kuwahara M, Teraoka A. Magnetic resonance imaging study of intracranial vertebrobasilar artery dissections. Stroke 1994; 25:571–575 8. Levy C, Laissy JP, Raveau V, et al. Carotid and vertebral artery dissections: three-dimensional time-of-flight MR angiography and MR imaging versus conventional angiography. Radiology 1994; 190:97–103 9. Zuber M, Meary E, Meder JF, Mas JL. Magnetic resonance imaging and dynamic CT scan in cervical artery dissections. Stroke 1994; 25:576–581 10. Ozdoba C, Sturzenegger M, Schroth G. Internal carotid artery dissection: MR imaging features and clinical–radiologic correlation. Radiology 1996; 199:191–198 11. Stringaris K, Liberopoulos K, Giaka E, et al. Three-dimensional time-of-flight MR angiography and MR imaging versus conventional angiography in carotid artery dissections. Int Angiol 1996; 15:20–25 12. Yoshimoto Y, Wakai S. Unruptured intracranial vertebral artery dissection: clinical course and serial radiographic imagings. Stroke 1997; 28:370– 374 13. Mascalchi M, Bianchi MC, Mangiafico S, et al. MRI and MR angiography of vertebral artery dissection. Neuroradiology 1997; 39:329–340 14. Auer A, Felber S, Schmidauer C, Waldenberger P, Aichner F. Magnetic resonance angiographic and clinical features of extracranial vertebral artery dissection. J Neurol Neurosurg Psychiatry 1998; 64:474–481 15. Leclerc X, Godefroy O, Salhi A, Lucas C, Leys D,

Pruvo JP. Helical CT for the diagnosis of extracranial internal carotid artery dissection. Stroke 1996; 27:461–466 16. Múnera F, Soto JA, Palacio D, Velez SM, Medina E. Diagnosis of arterial injuries caused by penetrating trauma to the neck: comparison of helical CT angiography and conventional angiography. Radiology 2000; 216:356–362 17. Chen CJ, Tseng YC, Lee TH, Hsu HL, See LC. Multisection CT angiography compared with catheter angiography in diagnosing vertebral artery dissection. Am J Neuroradiol 2004; 25:769– 774 18. Bub LD, Hollingworth W, Jarvik JG, Hallam DK. Screening for blunt cerebrovascular injury: evaluating the accuracy of multidetector computed tomographic angiography. J Trauma 2005; 59:691– 697 19. Schneidereit NP, Simons R, Nicolaou S. Utility of screening for blunt vascular neck injuries with computed tomographic angiography. J Trauma 2006; 60:209–216 20. Utter GH, Hollingworth W, Hallam DK, Jarvik JG, Jurkovich GJ. Sixteen-slice CT angiography in patients with suspected blunt carotid and vertebral artery injuries. J Am Coll Surg 2006; 203: 838–848 21. Eastman AL, Chason DP, Perez CL, McAnulty AL, Minei JP. Computed tomographic angiography for the diagnosis of blunt cervical vascular injury: is it ready for primetime? J Trauma 2006; 60:925–929 22. Malhotra AK, Camacho M, Ivatury RR, et al. Computed tomographic angiography for the diagnosis of blunt carotid/vertebral artery injury: a note of caution. Ann Surg 2007; 246:632–642 23. Pugliese F, Crusco F, Cardaioli G, et al. CT angiography versus colour Doppler US in acute dissection of the vertebral artery. Radiol Med 2007; 112:435–443 24. Sliker CW, Shanmuganathan K, Mirvis SE. Diagnosis of blunt cerebrovascular injuries with 16MDCT: accuracy of whole-body MDCT compared with neck MDCT angiography. AJR 2008; 190:790–799 25. Biffl WL, Ray CE Jr, Moore EE, Mestek M, Johnson JL, Burch JM. Noninvasive diagnosis of blunt cerebrovascular injuries: a preliminary report. J Trauma 2002; 53:850–856 26. Miller PR, Fabian TC, Croce MA, et al. Prospective screening for blunt cerebrovascular injuries: analysis of diagnostic modalities and outcomes. Ann Surg 2002; 236:386–395 27. Borisch I, Boehme T, Butz B, Hamer OW, Feuerbach S, Zorger N. Screening for carotid injury in trauma patients: image quality of 16-detector-row computed tomography angiography. Acta Radiol 2007; 48:798–805

1173

Downloaded from www.ajronline.org by 37.44.207.180 on 01/14/17 from IP address 37.44.207.180. Copyright ARRS. For personal use only; all rights reserved

Provenzale and Sarikaya 28. Inaba K, Munera F, McKenney M, et al. Prospective evaluation of screening multislice helical computed tomographic angiography in the initial evaluation of penetrating neck injuries. J Trauma 2006; 61:144–149 29. Berne JD, Reuland KS, Villarreal DH, McGovern TM, Rowe SA, Norwood SH. Sixteen-slice multidetector computed tomographic angiography im-

proves the accuracy of screening for blunt cerebrovascular injury. J Trauma 2006; 60:1204–1209 30. Wintermark M, Jawadi SS, Rapp JH, et al. Highresolution CT imaging of carotid artery atherosclerotic plaques. Am J Neuroradiol 2008; 29: 875–882 31. Elijovich L, Kazmi K, Gauvrit JY, Law M. The emerging role of multidetector row CT angiogra-

phy in the diagnosis of cervical arterial dissection: preliminary study. Neuroradiology 2006; 48:606–612 32. Vertinsky AT, Schwartz NE, Fischbein NJ, Rosenberg J, Albers GW, Zaharchuk G. Comparison of multidetector CT angiography and MR imaging of cervical artery dissection. Am J Neuroradiol 2008; 29:1753–1760

F O R YO U R I N F O R M AT I O N

Realize Your Potential with AJR Integrative Imaging Each quarterly issue of AJR Integrative Imaging offers you a series of peer-reviewed learning opportunities, many with CME credit, which will challenge and enhance your diagnostic and treatment skills. Whether it is the printed version or the online version, AJR Integrative Imaging is an important tool for your professional development. For more information or the latest issue of AJR Integrative Imaging, visit www.ajronline.org.

1174

AJR:193, October 2009