| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the University Department of Neurosurgery and Department of Neuroradiology, Institute of Neurological Sciences, Southern General Hospital, Glasgow, Scotland (P.M.W., E.M.T.); and the University of Edinburgh Departments of Clinical Neurosciences, Bramwell Dott Bldg, Western General Hospital, Crewe Rd, Edinburgh EH4 2XU, Scotland (P.M.W., J.M.W., V.E.), and Medical Statistics, Edinburgh, Scotland (V.E.). Received May 23, 2000; revision requested July 15; revision received August 21; accepted October 2. P.M.W. and V.E. supported by the British Brain and Spine Foundation from the Davie Cooper Scottish Aneurysm Study grant, administered by the University of Glasgow. J.M.W. supported by the Medical Research Council under the Clinical Research Initiative in Clinical Neuroscience, Medical Research Council, London, United Kingdom. Address correspondence to P.M.W. (e-mail: pmw@skull.dcn.ed.ac.uk).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: One hundred forty-two patients underwent intraarterial DSA to detect aneurysms. CT angiography, three-dimensional time-of-flight MR angiography, and intraarterial DSA were performed contemporaneously. Film hard-copy images and maximum intensity projection reconstructions of the CT angiograms and MR angiograms were reviewed at different times.
RESULTS: The accuracy per patient for the best observer was 0.87 at CT angiography and 0.85 at MR angiography. The accuracy per aneurysm for the best observer was 0.73 at CT angiography and 0.67 at MR angiography. Differences between readers and modalities were not significant. Interobserver agreement was good: 
value of 0.73 for CT angiography and of 0.74 for MR angiography. The sensitivity for detection of aneurysms smaller than 5 mm was 0.57 for CT angiography and 0.35 for MR angiography compared with 0.94 and 0.86, respectively, for detection of aneurysms 5 mm or larger. The accuracy of both CT angiography and MR angiography was lower for detection of internal carotid artery aneurysms compared with that at other sites. With low observer confidence, the likelihood of correct interpretation was significantly poorer.
CONCLUSION: CT angiography and MR angiography have limited sensitivity in the detection of small aneurysms but good interobserver agreement. There is no significant difference in diagnostic performance between the noninvasive modalities. Supplemental material: radiology.rsnajnls.org/cgi/content/full/219/3/739/DC1.
Index terms: Aneurysm, intracranial, 17.73 • Computed tomography (CT), angiography, 17.12112, 17.12115, 17.12116 • Digital subtraction angiography, comparative studies, 17.12483 • Magnetic resonance (MR), vascular studies, 17.121411, 17.121416, 17.12142
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
To compare imaging modalities directly, it is necessary to perform the examinations that are being evaluated in the same patient cohort. To our knowledge, however, only six studies have involved comparisons of both CT angiography and MR angiography with DSA in the same patients, and only two (sample sizes of 17 and 11 patients) of these six were prospective blinded studies (10,11). Of the four other studies, one focused on assessment of treated aneurysms rather than aneurysm detection (12). Two other studies by the same authors (13,14) were not prospective or blinded, and many patients in the second investigation did not undergo intraarterial DSA for comparison. The remaining study (15) also was not a prospective blinded investigation, and only five of 10 subjects underwent all three imaging examinations.
Although there is increasing interest in the use of CT angiography or MR angiography to replace intraarterial DSA as the primary angiographic modality in patients who present with SAH (16–19), perhaps the main use of CT angiography or MR angiography at present is to facilitate a diagnosis of aneurysm in asymptomatic at-risk patients or in patients who have symptoms that could have an aneurysmal origin but do not have acute SAH. Most previously published studies did not include such patients, so the accuracy in this type of population could not be truly determined. The purpose of our prospective blinded study was to compare CT angiography and MR angiography with intraarterial DSA for the detection of intracranial aneurysms in a large patient cohort.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
One hundred eighty-three consecutive patients who met the inclusion criteria and agreed to participate were recruited prospectively during an 18-month period. One hundred forty-seven of these patients (72 men, 75 women; median age, 41 years; age range, 19–75 years) underwent CT angiography, MR angiography, and intraarterial DSA. In 36 of 183 patients, only CT angiography or MR angiography could be performed in addition to intraarterial DSA, so these patients were excluded from further analysis. Five additional patients (one man and four women) were excluded because they were unable to complete the MR angiographic examination. Thus, our study population consisted of 142 patients.
Patients were grouped into four categories on the basis of the clinical indication for cerebral angiography. Group 1 consisted of four patients with one or more known aneurysms who underwent further assessment; group 2, 56 patients with proved SAH; group 3, 64 patients with symptoms that might be due to aneurysm; and group 4, 18 asymptomatic patients at risk of harboring an aneurysm. For subgroup analysis of accuracy according to clinical category, groups 1 and 2 were combined and groups 3 and 4 were combined.
The imaging studies were performed contemporaneously, if possible, and within a maximum of 2 months after intraarterial DSA. For CT angiography, 77 (54%) of 142 examinations were performed within 1 week after intraarterial DSA; an additional eight (6%) examinations, within 2 weeks; and the remaining 57 (40%) examinations, from 2 weeks to 2 months after intraarterial DSA. Reflecting the more limited access to MR imaging in the United Kingdom, fewer MR angiographic studies could be performed contemporaneously: 64 (45%) of 142 examinations were performed within 1 week after intraarterial DSA; an additional five (4%) examinations, within 2 weeks; and the remaining 73 (51%) examinations, from 2 weeks to 2 months after intraarterial DSA. Two MR angiographic studies were performed as part of a clinical MR imaging examination and because an aneurysm was suspected; intraarterial DSA was subsequently performed within 1 week. In all other cases, CT angiography and MR angiography were performed after intraarterial DSA.
Image Acquisition
Neuroradiologists (E.M.T., P.M.W.) performed all the intraarterial DSA examinations by using Advantx angiographic equipment (GE Medical Systems, Milwaukee, Wis). For intraarterial DSA, three- or four-vessel selective angiograms were acquired, with multiple projections obtained for each vessel. The images were printed onto film hard copy for analysis.
Spiral CT angiography was performed with a Twin scanner (Elscint, Haifa, Israel) in 81 examinations and with a HiSpeed Advantage scanner (GE Medical Systems) in 41 examinations. The acquisition volume was angled parallel to the superior orbitomeatal baseline, with the inferior margin at the superior surface of the posterior arch of the C1 vertebra (to include the posteroinferior cerebellar arteries) and extended superiorly to above the level of the pericallosal arteries. Tube angulation prevented irradiation of the orbits and allowed inclusion of all the usual aneurysm sites in the minimum examination volume. One hundred milliliters of nonionic contrast material (iopamidol 300; Bracco Diagnostics, Milan, Italy) was administered into an antecubital vein by using a pump injector at 3 mL/sec with an 18–20-second delay. The delay was increased when the patient was known to have substantial hemodynamic impairment; in such cases, a bolus tracking facility was used. CT examination parameters were 120-kV maximum tube current allowed, 512 x 512 matrix, 15-cm field of view, 1-mm collimation, and a pitch of 1.5 with a 0.5-mm reconstruction interval.
For logistic reasons—mainly scanner availability or patient illness—a small number of patients (n = 20) in Glasgow underwent nonspiral CT angiography with a 2400 Elite scanner (Elscint). Conventional transverse CT angiography was performed dynamically by using a 2.5-mm section width, 1-mm table increment, 120-kV 400-mAs tube current, and 20-cm field of view. Scanning began after 50 mL of iopamidol 300 was injected rapidly by hand; an additional 50 mL was injected during scanning (1).
MR angiography was performed by using a Prestige 2.0-T unit (Elscint) in 66 examinations and a Magnetom SP 1.5-T unit (Siemens, Erlangen, Germany) in 76 examinations, with three-dimensional time-of-flight MR angiographic sequences with magnetization transfer suppression and tilted optimized nonsaturating excitation, followed by a transverse T2-weighted fast spin-echo sequence. With the Prestige unit, two overlapping 58-mm slabs were acquired with the following parameters: 40/6 (repetition time msec/echo time msec), 30° flip angle, 15 x 20-cm field of view, 204 x 300 matrix, one signal acquired, and acquisition time of 7 minutes 53 seconds. With the Magnetom unit, one 64-mm slab angled 13° was obtained by using three-dimensional time-of-flight MR angiography. The parameters were 43/8, 20° flip angle, 20-cm field of view, 256 x 512 oversampled matrix, one signal acquired, and acquisition time of 11 minutes 48 seconds. Neuroradiologists supervised all CT angiographic and MR angiographic studies.
Postprocessing of Images
For CT angiography, reformatting of source images was performed on offline workstations (O2 Omnipro; Silicon Graphics, Mountain View, Calif, or Advantage Windows; GE Medical Systems) by a neuroradiology research fellow (P.M.W.) without review of the intraarterial DSA data prior to reformatting. Standard transverse and coronal oblique images plus curved sagittal multiplanar reformatted images were obtained by using the Omnipro workstation (1). Maximum intensity projection reconstructed angiograms also were obtained (12 projections at 15° intervals in both cranial-to-caudal and left-to-right projections), with bone editing performed by thresholding and manual cutting. Targeted maximum intensity projection imaging of the right and left internal carotid circulations and the vertebrobasilar system was performed. The total time for these reconstructions typically was 20–25 minutes.
With the Advantage Windows workstation, transverse, coronal oblique, and sagittal overlapping thick-slab (8–10 mm at 3–4-mm increments) maximum intensity projection images were obtained, with additional manual bone editing performed as required (2). The reconstruction time was less than 10 minutes unless manual bone editing was required. For MR angiography, standard maximum intensity projection reconstructed images were obtained at 15°–180° intervals (in cranial-to-caudal and left-to-right projections) at the time of MR angiography performed by the neuroradiographer. Targeted maximum intensity projection reconstructed images were obtained at examinations performed with the Prestige unit and Omnipro workstation, as in the previously described CT angiographic examination, again by the neuroradiology fellow and without review of the intraarterial DSA data in each case. Source CT angiograms and MR angiograms were available to the reviewers.
Image Review
Each patient was allocated a study number that was known only to the research fellow. The intraarterial DSA images were presented as film hard-copy images that were identified only by the study number—the patient names, examination dates, and clinician details were removed—in random order for independent review by two neuroradiologists (E.M.T., J.M.W.), one each from each center. Interpretation disagreements were resolved by means of consensus review. The CT angiograms and MR angiograms were presented in an anonymous random fashion to the same two neuroradiologists. The CT angiograms and MR angiograms obtained in any individual patient were reviewed separately, and at least 4 months elapsed between the review of DSA images and the review of the noninvasive studies in the same patient.
A number coding form was completed for each image so that all the major intracranial vessel segments were systematically reviewed in turn and an assessment was made as to whether the vessels were adequately demonstrated. The aneurysm sites and sizes were recorded. Site was recorded with the following codes: 1 for middle cerebral artery (MCA) mainstem, 2 for MCA bifurcation, 3 for distal MCA, 4 for anterior cerebral artery (ACA) complex, 5 for pericallosal artery, 6 for A1 ACA segment, 7 for internal carotid artery (ICA) bifurcation, 8 for posterior communicating artery, 9 for ophthalmic artery, 10 for ICA siphon, 11 for other ICA, 12 for basilar artery, 13 for posterior inferior cerebellar artery, and 14 for other artery (specified). For subgroup analysis, the sites were grouped into four categories: ACA complex, MCA complex, ICA complex (including posterior communicating artery), and vertebrobasilar system. Size was recorded as maximum angiographic dimension (a) smaller than 3 mm, (b) 3–5 mm, (c) 5.1–10.0 mm, or (d) larger than 10 mm.
Observer confidence was assessed by using a five-point scale, as reported by Atlas et al (20), on which a score of 5 meant aneurysm definitely absent; 4, aneurysm probably absent; 3, uncertain; 2, aneurysm probably present; and 1, aneurysm definitely present. In a multicenter multimodality study involving several types of imaging equipment, a large number of variables are introduced. We studied the effect of several variables that have been suggested to be relevant to diagnostic accuracy (21), including the use of spiral versus nonspiral CT angiography, the availability of spin-echo sequences for MR angiography, and the availability of targeted maximum intensity projection reconstruction for CT angiography and MR angiography.
Statistical Analyses
For both modalities and both observers, 2 x 2 tables of the true-positive, false-positive, true-negative, and false-negative cases at MR angiography and CT angiography, as compared with those at intraarterial DSA, were constructed. Sensitivity, specificity, positive and negative predictive values, and accuracy were compared on a per-patient (ie, the ability to correctly designate a patient as a true-positive or true-negative case on the basis of possession of at least one intracranial aneurysm) and per-aneurysm (ie, the ability to correctly identify all aneurysms) basis. Exact 95% CIs based on binomial probabilities were calculated (22). The McNemar test for paired binary data was used to compare the sensitivity and specificity of CT angiography and MR angiography with each observer (23). The unweighted statistic was used to assess interobserver and intermodality agreement (24). A value of 0.8 or above indicated excellent agreement; 0.6–0.8, good agreement; 0.4–0.6, fair agreement; and less than 0.4, poor agreement. CIs for the difference between two proportions were calculated to determine whether there was a difference between the proportions interpreted correctly for each modality and each observer at different levels of observer confidence (23).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Details of all the false-negative and false-positive results of CT angiography and MR angiography are given in Table E2 (radiology.rsnajnls.org/cgi/content/full/219/3/739/DC1). Observer A made 33 false-negative readings of aneurysms at CT angiography and 52 at MR angiography, compared with observer B, who made 36 and 54 false-negative readings, respectively. When the two readers’ mistakes were combined, 28 (41%) of the 69 false-negative readings at CT angiography were related to aneurysms smaller than 3 mm; 37 (54%), to aneurysms 3–5 mm; four (6%), to aneurysms 5.1–10.0 mm; and none, to aneurysms larger than 10 mm. Six (9%) of the 69 false-negative CT angiographic readings were related to ACA aneurysms; 16 (23%), to MCA aneurysms; 36 (52%), to ICA aneurysms; and 11 (16%), to vertebrobasilar aneurysms. At MR angiography, 36 (34%) of the 106 false-negative readings were related to aneurysms smaller than 3 mm; 61 (58%), to aneurysms 3–5 mm; six (6%), to aneurysms 5.1–10.0 mm; and three (3%), to aneurysms larger than 10 mm. Nineteen (18%) of the 106 false-negative MR angiographic readings were related to ACA aneurysms; 28 (26%), to MCA aneurysms; 46 (43%), to ICA aneurysms; and 13 (12%), to vertebrobasilar aneurysms.
Observer A made 26 false-positive readings at CT angiography and 11 at MR angiography compared with observer B, who made 18 and 11 false-positive readings, respectively. When the two observers’ readings were combined, at CT angiography, 28 (64%) of 44 false-positive aneurysms were smaller than 3 mm; 14 (32%), 3–5 mm; two (4%), 5.1–10.0 mm; and none, larger than 10 mm. Eight (18%) of these 44 findings were categorized as ACA aneurysms; 17 (39%), MCA aneurysms; 15 (34%), ICA aneurysms; and four (9%), vertebrobasilar aneurysms. At MR angiography, 15 (68%) of 22 false-positive aneurysms were smaller than 3 mm; five (23%), 3–5 mm; one (4%), 5.1–10.0 mm; and one (4%), larger than 10 mm. One (4%) of these 22 false-positive findings was an ACA aneurysm; eight (36%), MCA aneurysms; 11 (50%), ICA aneurysms; and two (9%), vertebrobasilar aneurysms.
The overall comparative diagnostic performances of CT angiography and MR angiography (with 95% CIs) for both observers are listed in Table 1 and illustrated graphically as Forrest plots in Figure 1 (25). The accuracy of both CT angiography and MR angiography on a per-patient basis was better than the accuracy on a per-aneurysm basis. There was no significant difference in sensitivity between CT angiography and MR angiography (P = .11 for observer A and P = .10 for observer B, McNemar test). For observer A only (P = .01), CT angiography had significantly poorer specificity than did MR angiography (P = .56 for observer B). Agreement between the noninvasive modalities was good: the statistic for observer A was 0.61 (95% CI: 0.48, 0.74) and for observer B, 0.69 (95% CI: 0.57, 0.81). Interobserver agreement was good for both noninvasive modalities, with a statistic of 0.73 (95% CI: 0.62, 0.84) for CT angiography and of 0.74 (95% CI: 0.63, 0.86) for MR angiography (24).
|
|
|
|
|
|
|
|
|
|
|
|
Because clinically the most important categories of patients were those in groups 1 and 2—because they were much more likely to have aneurysms that required treatment—an assessment was made also to determine whether the clinical group influenced diagnostic performance by site and aneurysm size (ie, groups 1 and 2 versus groups 3 and 4). These data are presented in Tables 2 and 3 and indicate that although diagnostic accuracy was generally better in groups 3 and 4 combined, the differences were generally small, with widely overlapping 95% CIs between the clinical groups. The only exceptions were the diagnostic accuracy at MR angiography of aneurysms 3–5 mm for both observers and that at MR angiography of MCA complex aneurysms for observer B: Accuracy was substantially greater in groups 3 and 4 than in groups 1 and 2.
For patients in groups 1 and 2 combined, the accuracies per patient for CT angiography were 0.82 (49 of 60) (95% CI: 0.7, 0.9) and 0.85 (51 of 60) (95% CI: 0.73, 0.93) for observers A and B, respectively. For patients in groups 3 and 4 combined, the accuracies per patient for CT angiography were 0.83 (68 of 82) (95% CI: 0.73, 0.90) and 0.89 (73 of 82) (95% CI: 0.80, 0.95) for observers A and B, respectively. For MR angiography, the accuracies per patient for groups 1 and 2 were 0.82 (49 of 60) (95% CI: 0.7, 0.9) and 0.83 (50 of 60) (95% CI: 0.71, 0.92) for observers A and B, respectively; for groups 3 and 4, this value was 0.87 (71 of 82) (95% CI: 0.77, 0.93) for both observers.
When observers were confident that an aneurysm was present or absent (confidence score 1 or 5, respectively, on a five-point scale), the proportion of cases interpreted correctly (ie, as true-positive or true-negative) at CT angiography was 0.92 (85 of 92) for observer A and 0.78 (124 of 159) for observer B. At MR angiography, these proportions were 0.84 (99 of 118) and 0.74 (114 of 154), respectively. When the observers believed that an aneurysm was probably present or absent (confidence score of 2 or and 4, respectively), the proportion of cases interpreted correctly at CT angiography was 0.60 (46 of 76) for observer A and 0.50 (nine of 18) for observer B; at MR angiography, these values were 0.44 (27 of 61) and 0.36 (10 of 28), respectively. When the observers were uncertain about the presence or absence of an aneurysm (score 3), the proportion of cases interpreted correctly at CT angiography was 0.29 (nine of 31) for observer A and 0.52 (11 of 21) for observer B; for MR angiography, these values were 0.29 (four of 14) and 0.36 (four of 11), respectively.
The difference in proportion of studies correctly interpreted (with 95% CI) in the uncertain (confidence score 3) category was compared with this proportion in the combined definite and probable categories (confidence scores 1, 2, 4, and 5). For both observers and at both CT angiography and MR angiography, the difference in proportion interpreted correctly was significantly poorer for the uncertain category compared with all other categories (P <. 05).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
There are several possible reasons for this discrepancy, including the aneurysm size distribution, prospective study design, lower prevalence of aneurysms, intention to image analysis, complete blinding of image review, and use of film hard-copy images alone for review in this study. In addition, there may be a trend in technology assessment toward small, early publications, with highly selected populations producing more optimistic results than later larger studies. For the purposes of the study, we deliberately used widely available and established clinical imaging protocols rather than rapidly evolving "cutting-edge" techniques such as fast gradient-echo contrast-enhanced MR angiography. The fact that the equipment used for the current study is now relatively out of date and produces lower quality images relative to advanced technology did not affect the comparison of our study data with those in the literature, because equivalent technology was used for virtually all the studies published to the end of 1998. Such equipment will be replaced gradually and continue in use for some time in many parts of the world; therefore, these data are relevant to many radiologists.
Time-of-flight MR angiography does not depict some aneurysms. This is mainly because of spin saturation secondary to slow flow and/or phase dispersion effects due to turbulent flow in an aneurysm (29). In such turbulent or slow flow areas, rapid gradient-echo techniques with very short repetition and echo times and gadolinium-enhanced imaging enable flow-independent imaging of blood, greatly reduce turbulence artifacts, and eliminate in-plane saturation effects (30–32). Such techniques have been used in recent years in studies of the aorta, cervical arteries, and other aortic branches. Owing to the problems of data acquisition timing to obtain precise filling of central k space (the low spatial frequency data) at the time of maximal arterial contrast material concentration in the vascular bed of interest, contrast-enhanced MR angiography has not always proved superior to time-of-flight MR angiography (33). Current technical developments such as fluoroscopic triggering or temporally resolved contrast-enhanced MR angiography may solve this problem and thereby facilitate the more general clinical use of contrast-enhanced MR angiography (34).
Surprisingly, until recently, little information was available on the in vivo clinical use of ultrafast contrast-enhanced MR angiography for the detection of intracranial aneurysms (35). In a very recent (to our knowledge, the first) prospective study (35) involving 32 patients, 17 of whom had a total of 23 aneurysms, contrast-enhanced MR angiography had greater sensitivity than did time-of-flight MR angiography (100% vs 96%) but poorer specificity (94% vs 100%); both the sensitivity and specificity with the state-of-the-art MR imaging unit used in this small series were impressive. Results of a recent in vitro study (29) of an aneurysm model also indicated the advantage of ultrafast contrast-enhanced MR angiography over time-of-flight MR angiography. Multisection CT with submillimeter collimation, which offers improved spatial resolution and reduced overlap from venous structures as a result of rapid scanning, may further improve CT angiography accuracy, although prospective clinical studies have not yet been performed to confirm this.
In this study, 72 (67%) of 108 aneurysms were small (
5 mm). Small aneurysms are much harder to detect than are larger aneurysms; for example, a sensitivity of 25% for detection of aneurysms smaller than 3 mm versus 92% for detection of larger aneurysms (36) and an accuracy for detection of small aneurysms as low as 0.56 (37) have been reported at MR angiography. Many previous studies have not provided detailed information on aneurysm size, and those that have had a greater proportion of aneurysms larger than 5 mm. Nearly all of the false-negative results in the present study were for aneurysms 5 mm or smaller and were proportionately concentrated among those smaller than 3 mm. The diagnostic performance of CT angiography and MR angiography for detection of aneurysms larger than 5 mm was excellent in patients with acute SAH or known aneurysm and those without these abnormalities (Table 2). Although aneurysms larger than 5 mm comprised 33% (36 of 108) of the aneurysms in this series, they accounted for only 6% (four of 69 at CT angiography) to 9% (10 of 106 at MR angiography) of the false-negative readings and 4% (two of 44 at CT angiography) to 9% (two of 22 at MR angiography) of the false-positive readings. Conversely, very small aneurysms (<3 mm) comprised 22% (22 of 108) of all aneurysms, yet they accounted for 34% (36 of 106 at MR angiography) to 41% (28 of 69 at CT angiography) of the false-negative readings and 64% (28 of 44 at CT angiography) to 68% (15 of 22 at MR angiography) of the false-positive readings.
The population in the present study may have contributed to the poorer results per aneurysm than those previously reported. We sought to recruit patients who did not have a known aneurysm or acute SAH (as well as acute SAH cases), because one cannot assume that the imaging results from previous studies in a high aneurysm prevalence population will necessarily be the same in lower prevalence population groups. In the present study, 60 (42%) of 142 patients had a known aneurysm or proved SAH, and the overall aneurysm prevalence was 63 (44%) of 142 patients. This is low compared with the results in most of the previous studies, in which the aneurysm prevalence was 75% or greater in the majority of cases (2,6,8–11,17,18,20,27,38–44). There is evidence that increasing disease prevalence can lead to an improvement in the sensitivity and specificity of an examination, whereas it had previously been thought that increasing prevalence only influenced the predictive values (45).
If nearly all the patients in a study are known to have an aneurysm or SAH, this could lead to observer expectation bias. The distribution of subarachnoid blood may provide a strong clue to the presence and/or site of an aneurysm, as may a local hematoma or the presence of hydrocephalus. However, this advantage may be offset by the potential to obtain poorer quality images in sick, restless patients with acute SAH: The long acquisition time of three-dimensional time-of-flight MR angiography makes these patients particularly susceptible to this problem (Fig 2). Our study data support this conclusion, because the diagnostic performance of CT angiography and MR angiography was consistently slightly better in clinical groups 3 and 4 than in groups 1 and 2 (Tables 2 and 3). Asymptomatic patients are more likely to have small aneurysms compared with patients who have an SAH. Aneurysms 5 mm or smaller accounted for one-third of all the aneurysms in one large study (46) involving asymptomatic patients with unruptured aneurysms, and they accounted for 72 (67%) of the 108 aneurysms in the current study. As expected, in group 2 (with recent SAH), aneurysm prevalence was greater and the aneurysms were on average larger: 21 (40%) of 53 aneurysms were smaller than 5 mm compared with 10 (22%) of 45 aneurysms in groups 3 and 4.
In clinical practice, which we attempted to reproduce in this study, a single reader of a diagnostic imaging study is usual. Some earlier studies (10,11,15–17,26,28, 39,42–44,47–53) involved consensus review by two or more readers in the analysis of accuracy; this could result in better accuracy and therefore a positive bias toward the noninvasive methods. It was reassuring to find that in a large prospective blinded study, interobserver agreement was good for both CT angiography and MR angiography. It was somewhat surprising, however, that we did not find any advantage in using targeted maximum intensity projection reconstructions at either CT angiography or MR angiography (Table 4). One potential limitation of this study was the delay in some cases between intraarterial DSA and CT angiography or MR angiography (see Materials and Methods). A delay of several weeks between intraarterial DSA and the noninvasive study could result in an aneurysm clotting or not being seen at intraarterial DSA recanalization. This could result in a false-negative or false-positive result for the noninvasive examination when in fact it was a true-negative or true-positive study. In the current study, however, this effect appeared to be very small, as indicated by the results of a comparison of contemporaneous CT angiography and MR angiography with delayed examinations.
In the small subgroup of 18 asymptomatic patients, 12 had a total of 21 aneurysms, three of which were larger than 5 mm. Probably reflecting this size distribution, CT angiography and MR angiography performed poorly in this patient subgroup, with mean sensitivities per patient of 0.67 (16 of 24 at CT angiography) and 0.55 (23 of 42 at MR angiography) and mean accuracies per patient of 0.75 (27 of 36 at CT angiography) and 0.69 (25 of 36 at MR angiography). Therefore, in a low-prevalence asymptomatic population (eg, one undergoing noninvasive examinations for aneurysm screening), one can expect considerably poorer diagnostic performance. The accuracy per patient was 12% (at CT angiography) to 27% (at MR angiography) lower in the "screening" subgroup (group 4) than that in the other subgroups combined: 0.72 (13 of 18) versus 0.84 (104 of 124) at CT angiography and 0.61 (11 of 18) versus 0.88 (109 of 124) at MR angiography for the observer with the largest such effect (observer A).
A number of important lessons for the use of noninvasive examinations for aneurysm detection were highlighted in this study. First, diagnostic performance is substantially limited by aneurysm size. Second, at certain sites, particularly those where there is considerable vessel overlap or adjacent bone (such as the cavernous and terminal ICA segments or the MCA bifurcation), both CT angiography and MR angiography perform more poorly. A higher than expected number of false-positive readings were related to very small MCA aneurysms, and a higher than expected proportion of false-negative readings were related to ICA aneurysms. Therefore, caution should be exercised in the interpretation of small aneurysms arising from the MCA bifurcation or the intracranial ICA and its branches at CT angiography or MR angiography, particularly if observer confidence is low.
In light of these data and those in the literature (25), we believe that current standard clinical CT angiography and MR angiography cannot yet safely replace intraarterial DSA in the diagnostic work-up of patients with acute SAH because of the low sensitivity of these modalities in the detection of small aneurysms; however, these examinations may be useful adjuncts (26). It is well recognized that small aneurysms can rupture, and the potential adverse consequences to the patient of missing such an aneurysm at CT angiography and/or MR angiography performed instead of intraarterial DSA are severe. In this clinical setting, a false-positive result would have less serious consequences, provided that confirmatory intraarterial DSA was performed before aneurysm treatment in all positive cases. For patients with symptoms that are strongly suggestive of an aneurysm, we believe that although both CT angiography and MR angiography can help exclude the presence of an aneurysm larger than 5 mm (ie, one likely to be large enough to cause compressive symptoms) fairly reliably, intraarterial DSA should be considered the investigation of choice, because a negative CT angiogram or MR angiogram cannot completely ensure the diagnosis, and a positive study will lead to intraarterial DSA.
When clinical suspicion of aneurysm as the cause of the symptoms is low, a noninvasive examination alone is adequate, provided both the patient and clinician can accept the uncertainty that a small aneurysm (5 mm) has not been excluded. In light of the International Study of Unruptured Intracranial Aneurysms Investigators data (54), only aneurysms larger than 10 mm should be considered for treatment in asymptomatic patients with no prior history of SAH. Therefore, at-risk patients may be adequately examined by using CT angiography or MR angiography alone, but again, only if both the patient and the clinician can accept the uncertainty about small aneurysms. We emphasize that according to the best available evidence, the examination of asymptomatic at-risk patients is not routinely indicated (5).
In conclusion, CT angiography and MR angiography are equally accurate in the detection of intracranial aneurysms, and aneurysms 5 mm or smaller are detected substantially less well than are those larger than 5 mm. Accuracy on a per-patient basis is better than that on a per-aneurysm basis. Interobserver agreement is good for both modalities. Using a simple confidence scoring system is a useful means of determining individual cases in which the noninvasive examination is likely to be less reliable. In a screening situation, accuracy is expected to be lower than the overall per-patient accuracy level of 0.85 that was achieved in this study.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, P.M.W., J.M.W.; study concepts and design, E.M.T., J.M.W., P.M.W.; literature research, P.M.W.; clinical studies, P.M.W., E.M.T., J.M.W.; data acquisition, E.M.T., P.M.W., J.M.W.; data analysis/interpretation, P.M.W., V.E.; statistical analysis, V.E.; manuscript preparation, P.M.W.; manuscript definition of intellectual content, all authors; manuscript editing, J.M.W., E.M.T., V.E.; manuscript revision/review, all authors; manuscript final version approval, P.M.W., J.M.W., E.M.T.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  Q. Li, F. Lv, Y. Li, T. Luo, K. Li, and P. Xie Evaluation of 64-Section CT Angiography for Detection and Treatment Planning of Intracranial Aneurysms by Using DSA and Surgical Findings Radiology, June 9, 2009; (2009) 2523081911. [Abstract] [Full Text]
|
|
|
|
 |
|
 J. D. Burns, J. Huston III, K. F. Layton, D. G. Piepgras, and R. D. Brown Jr Intracranial Aneurysm Enlargement on Serial Magnetic Resonance Angiography: Frequency and Risk Factors Stroke, February 1, 2009; 40(2): 406 - 411. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Koffijberg, G. Rinkel, and E. Buskens Do Intraindividual Variation in Disease Progression and the Ensuing Tight Window of Opportunity Affect Estimation of Screening Benefits? Med Decis Making, January 1, 2009; 29(1): 82 - 90. [Abstract] [PDF]
|
|
|
|
 |
|
 T. J. Kaufmann and D. F. Kallmes Diagnostic Cerebral Angiography: Archaic and Complication-Prone or Here to Stay for Another 80 Years? Am. J. Roentgenol., June 1, 2008; 190(6): 1435 - 1437. [Full Text] [PDF]
|
|
|
|
|
|
A. Kemmling, I. Noelte, L. Gerigk, S. Singer, C. Groden, and J. Scharf A Diagnostic Pitfall for Intracranial Aneurysms in Time-of-Flight MR Angiography: Small Intracranial Lipomas Am. J. Roentgenol., January 1, 2008; 190(1): W62 - W67. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Lubicz, M. Levivier, O. Francois, P. Thoma, N. Sadeghi, L. Collignon, and D. Baleriaux Sixty-Four-Row Multisection CT Angiography for Detection and Evaluation of Ruptured Intracranial Aneurysms: Interobserver and Intertechnique Reproducibility AJNR Am. J. Neuroradiol., November 1, 2007; 28(10): 1949 - 1955. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. C. Bowen and C. L. Truwit MR Angiography versus CT Angiography in the Evaluation of Neurovascular Disease Radiology, November 1, 2007; 245(2): 357 - 361. [Full Text] [PDF]
|
|
|
|
|
|
K. Papke, C. K. Kuhl, M. Fruth, C. Haupt, M. Schlunz-Hendann, D. Sauner, M. Fiebich, A. Bani, and F. Brassel Intracranial Aneurysms: Role of Multidetector CT Angiography in Diagnosis and Endovascular Therapy Planning Radiology, August 1, 2007; 244(2): 532 - 540. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D.Y. Yoon, K.J. Lim, C.S. Choi, B.M. Cho, S.M. Oh, and S.K. Chang Detection and Characterization of Intracranial Aneurysms with 16-Channel Multidetector Row CT Angiography: A Prospective Comparison of Volume-Rendered Images and Digital Subtraction Angiography AJNR Am. J. Neuroradiol., January 1, 2007; 28(1): 60 - 67. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. L. Brisman, J. K. Song, and D. W. Newell Cerebral aneurysms. N. Engl. J. Med., August 31, 2006; 355(9): 928 - 939. [Full Text] [PDF]
|
|
|
|
 |
|
 G. M. Teasdale, J. M. Wardlaw, P. M. White, G. Murray, E. M. Teasdale, V. Easton, and on behalf of the Davie Cooper Scottish Aneurysm St The familial risk of subarachnoid haemorrhage Brain, July 1, 2005; 128(7): 1677 - 1685. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. M. Baddour, W. R. Wilson, A. S. Bayer, V. G. Fowler Jr, A. F. Bolger, M. E. Levison, P. Ferrieri, M. A. Gerber, L. Y. Tani, M. H. Gewitz, et al. Infective Endocarditis: Diagnosis, Antimicrobial Therapy, and Management of Complications: A Statement for Healthcare Professionals From the Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease, Council on Cardiovascular Disease in the Young, and the Councils on Clinical Cardiology, Stroke, and Cardiovascular Surgery and Anesthesia, American Heart Association: Endorsed by the Infectious Diseases Society of America Circulation, June 14, 2005; 111(23): e394 - e434. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K Doi Current status and future potential of computer-aided diagnosis in medical imaging Br. J. Radiol., January 1, 2005; 78(suppl_1): S3 - s19. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Bendszus, M. Koltzenburg, A. J. Bartsch, R. Goldbrunner, T. Gunthner-Lengsfeld, F. X. Weilbach, K. Roosen, K. V. Toyka, and L. Solymosi Heparin and Air Filters Reduce Embolic Events Caused by Intra-Arterial Cerebral Angiography: A Prospective, Randomized Trial Circulation, October 12, 2004; 110(15): 2210 - 2215. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Teksam, A. McKinney, S. Casey, M. Asis, S. Kieffer, and C. L. Truwit Multi-Section CT Angiography for Detection of Cerebral Aneurysms AJNR Am. J. Neuroradiol., October 1, 2004; 25(9): 1485 - 1492. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. A. Willinek, M. Born, B. Simon, H. J. Tschampa, C. Krautmacher, J. Gieseke, H. Urbach, H. J. Textor, and H. H. Schild Time-of-Flight MR Angiography: Comparison of 3.0-T Imaging and 1.5-T Imaging--Initial Experience Radiology, December 1, 2003; 229(3): 913 - 920. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Vogel, R. Verreault, J.-F. Turcotte, M. Kiesmann, and M. Berthel Review Article. Intracerebral Aneurysms: A Review With Special Attention to Geriatric Aspects J. Gerontol. A Biol. Sci. Med. Sci., June 1, 2003; 58(6): M520 - 524. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. A. Willinsky, S. M. Taylor, K. terBrugge, R. I. Farb, G. Tomlinson, and W. Montanera Neurologic Complications of Cerebral Angiography: Prospective Analysis of 2,899 Procedures and Review of the Literature Radiology, May 1, 2003; 227(2): 522 - 528. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. F. Tomandl, P. Hastreiter, S. Iserhardt-Bauer, N. C. Kostner, M. Schempershofe, W. J. Huk, T. Ertl, C. Strauss, and J. Romstock Standardized Evaluation of CT Angiography with Remote Generation of 3D Video Sequences for the Detection of Intracranial Aneurysms RadioGraphics, March 1, 2003; 23(2): e12 - e12. [Abstract] [Full Text]
|
|
|
|
|
|
A. Mallouhi, S. Felber, A. Chemelli, A. Dessl, A. Auer, M. Schocke, W. R. Jaschke, and P. Waldenberger Detection and Characterization of Intracranial Aneurysms with MR Angiography: Comparison of Volume-Rendering and Maximum-Intensity-Projection Algorithms Am. J. Roentgenol., January 1, 2003; 180(1): 55 - 64. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Safriel, J. Nath, J. M. Wardlaw, and P. White Noninvasive Methods in the Detection of Intracranial Aneurysms * Drs Wardlaw and White respond: Radiology, June 1, 2002; 223(3): 881 - 882. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
