| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the Department of Radiology, Kumamoto University School of Medicine, 1-1-1, Honjo, Kumamoto City, 860 Japan (Y.K., M.T.); the Department of Radiology, School of Medicine, Fujita Health University, Fujita, Japan (K.K., Y.O.); the Department of Radiology, Faculty of Medicine, Kyushu University, Fukuoka, Japan (K.H.); the Department of Radiology, Nagasaki University School of Medicine, Nagasaki, Japan (M.O.); the Department of Radiology, Fukuoka University School of Medicine, Fukuoka, Japan (H.U.); the Department of Radiology, Kurume University School of Medicine, Kurume, Japan (T.A.); and the Department of Radiology, National Cardiovascular Center, Osaka, Japan (S.I.). From the 1997 RSNA scientific assembly. Received January 22, 1998; revision requested March 6; final revision received August 31; accepted October 26. Address reprint requests to Y.K.
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: The 3D CT angiograms obtained in 49 patients with or without intracranial aneurysms were evaluated by four blinded observers. Results were compared with findings at conventional angiography or surgery. A volume-rendering method was used, and 13 images obtained in different directions were reviewed in each study. The diameter of aneurysms was divided into four sizes: large, greater than 13 mm; medium, 5–12 mm; small, 3–4 mm; and very small, less than 3 mm. Results were also evaluated by means of receiver operating characteristic analysis.
RESULTS: At conventional angiography, 47 aneurysms, including 14 less than 3 mm, were depicted in 35 patients. The mean sensitivity of CT angiography for very small aneurysms was 64%; small, 83%; medium, 95%; and large, 100%. Some very small aneurysms that were not depicted at conventional angiography were depicted at CT angiography, and one was proved at surgery.
CONCLUSION: CT angiography has good sensitivity for depiction of intracranial aneurysms 3 mm or larger and relatively good sensitivity for aneurysms less than 3 mm. CT angiography may be a noninvasive technique for detection of asymptomatic unruptured or ruptured aneurysms.
Index terms: Aneurysm, CT, 17.12115, 17.12116, 17.12117, 17.73 • Aneurysm, intracranial, 17.73 • Aneurysm, MR, 17.12142, 17.73 • Cerebral blood vessels, CT, 17.12115, 17.12116, 17.12117 • Cerebral blood vessels, MR, 17.12142 • Computed tomography (CT), three-dimensional, 17.12117
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
In some patients, however, MR angiography may be prohibited owing to the presence of indwelling electric devices or ferromagnetic intracranial aneurysm clips and may be limited in others as a result of motion artifact. Helical CT is a recently developed technique in which scanning is performed while the CT table is drawn through the gantry, allowing a continuous volume of transaxial data to be rapidly generated (2,3). With this technique, vascular structures can be selectively imaged with an appropriate delay after intravenous injection of contrast material and can be reconstructed to produce a three-dimensional (3D) representation. In one study, aneurysms 3 mm or larger were seen clearly with helical CT, but those smaller than 3 mm were not detected (2).
Many authors have reported the clinical usefulness of CT angiography with maximum intensity projection (MIP) and shaded surface display (2–9). Volume rendering, a third method for generating 3D images from CT data sets, uses information from all voxels within a volume, avoiding the extensive loss of information that is inherent in MIP and shaded surface display (10,11). The volume-rendering technique also provides substantially greater control over the display of structures with varying attenuation than is provided with MIP and shaded surface display. To our knowledge, however, no large studies have systematically evaluated the usefulness of CT angiography with volume-rendering technique for detection of intracranial aneurysms. Also, although at least 10% of subarachnoid hemorrhage occurs in the posterior fossa, many previous studies involved only the circle of Willis.
The purpose of this study was to assess the diagnostic accuracy of 3D CT angiography with volume rendering for the detection of intracranial aneurysms, especially very small aneurysms (diameter less than 3 mm).
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
At conventional angiography, aneurysms were not seen in four of the 23 patients with subarachnoid hemorrhage; in three of the 14 patients in whom aneurysms had been suspected at CT or MR imaging, MR angiography, or both; and in seven of the 12 patients with a variety of diseases. Therefore, aneurysms were seen at conventional angiography in a total of 35 study patients (11 men and 24 women; average age, 60.7 years; age range, 35–82 years) (19 with subarachnoid hemorrhage; 11 with aneurysm suspected at CT or MR imaging, MR angiography, or both; and five with other diseases). Aneurysms were not seen at conventional angiography in a total of 14 patients (six men and eight women; average age, 53.9 years; age range, 35–76 years) (four with subarachnoid hemorrhage; three with aneurysm suspected at CT or MR imaging, MR angiography, or both; and seven with other diseases). These 14 patients served as the control subjects in this blinded study.
Institutional review board approval was obtained for the study, and each study patient and control subject gave written informed consent.
In all study patients and control subjects, conventional four-vessel cerebral angiography was performed with the Seldinger method via the femoral artery by means of either screen-film or digital subtraction technique. Anteroposterior, lateral, and oblique views were obtained with stereoscopic technique. We calculated the size of each cerebral aneurysm after correction for the magnification factor. The diameter of each aneurysm was graded as large (greater than 13 mm), medium (5–12 mm), small (3–4 mm), or very small (less than 3 mm). The presence or absence of aneurysms, as well as the location and the size when present, was established by means of consensus by two experienced reviewers (K.K., Y.O.), who did not participate in interpretation of the CT angiograms. The conventional angiograms were used as the standard of reference.
Scanning Techniques and Production of 3D Images
The scanning parameters for helical CT (Xpress/SX; Toshiba, Tokyo, Japan) included 135 kV, 220 mA, section thickness of 0.8 or 1 mm, and bed speed of 0.8 or 1 mm/sec (12,13). The scanning time for one rotation was 1 second, and the total scanning time was 30 seconds. Spiral data for CT angiography were obtained by means of inferosuperior scanning. After a starting point at the foramen magnum was selected, 80–100 mL of nonionic contrast material (iodine concentration, 300 mg/mL) (iopamidol, Iopamiron; Nihon Schering, Osaka, Japan) was injected intravenously at a rate of 2.0–3.0 mL/sec by using a power injector (Medrad, Pittsburgh, Pa) via an 18- or 20-gauge needle inserted in the antecubital vein.
Scanning was started by using a triggering technique to optimize acquisition of early arterial phase images based on the arrival of contrast material into the region of interest. An application of real-time CT (SureStart function) was used to trigger scanning, and the region of interest was placed at the internal carotid artery (ICA) in the carotid canal. With use of the reconstruction program algorithm of a 180° opposite beam interpolation, 60–70 axial images with 0.4- or 0.5-mm pitch were generated from the raw data. This allowed visualization from the vertebral artery proximal to the branch of the posterior inferior cerebellar artery to the distal anterior and middle cerebral arteries (just above the A2–A3 junction).
The CT data set for each case was transferred to a freestanding workstation. The 3D CT reformations were obtained by means of the volume-rendering method of the computer analyzing system for digital images. Radiologists performed the reconstructions of 3D images. The thresholding technique had a lower threshold of 95–110 HU. Bone structures were not edited out of the axial images. With use of the triggering technique and real-time CT, the increased attenuation of acute subarachnoid blood had no effect on the ability to define the vessels during 3D reconstruction, since the attenuation of cisternal blood was lower than that of the enhanced vessels in all cases. Reformation of one 3D CT image took 2 minutes (30–40 minutes per case).
With use of 3D CT angiography, we obtained a total of 13 color images, each at a different angle. The size of each image was 18 x 10, 18 x 13, or 18 x 18 cm depending on coverage of the intracranial vasculature. The different angles included a stereoscopic view from above, a view from below, a stereoscopic view from the front (with slight superoinferior angulation of approximately 10°), a view from behind (with superoinferior angulation of approximately 30°), a view from behind after the posterior circulation was edited out, lateral stereoscopic views from the right and left, median views from the right (with the right side edited out) and left (with the left side edited out). To illustrate desired features, voxel opacity settings, color, lighting, and viewing direction were adjusted for each data set. Final image adjustments for printing were performed with PHOTOSHOP 3.03 (Adobe Systems, Mountain View, Calif).
Reading Methods
The CT angiograms were reviewed retrospectively by four trained observers (K.H., M.O., H.U., T.A.), who were blinded to results at conventional angiography and surgery. The images were randomized, and the order in which the images were reviewed was varied systematically among observers as follows: observer A began interpreting CT angiograms from case 1, observer B from case 13, observer C from case 25, and observer D from case 37. The CT angiograms were graded by means of a continuous confidence-judgment scale (14). The continuous scale allowed a subjective probability estimate (ranging from 0%–100%) that an aneurysm was present. The location and size of the aneurysms were also evaluated.
After the blinded study, the CT angiograms of all 49 patients were reviewed retrospectively by the same four readers with the findings at conventional angiography and surgery. Aneurysms missed by three or more observers were reassessed to analyze the reasons why they were missed. To evaluate which scanning directions were useful, the depiction of aneurysms on all images was classified into four grades: grade 3, whole aneurysm depicted clearly; grade 2, shape of aneurysm partially unclear; grade 1, aneurysm only partially depicted; and grade 0, aneurysm not seen.
Data Analysis
Receiver operating characteristic (ROC) analysis was performed to assess the ability to detect aneurysms on CT angiograms (15). In this evaluation, detection of at least one aneurysm was assumed positive for multiple aneurysms; that is, the diagnosis of any one aneurysm would lead to conventional angiography and, thus, the detection of any additional aneurysms. Data from the four observers were not pooled; hence, an ROC curve was calculated for each observer. The ROC curves were fitted to the data points with the LABROC-1 software (15,16), which also calculates the area under the curve, or Az (17).
In the continuous confidence-judgment scale (ranging from 0%–100%), ratings of 50% and 75% were chosen as cutoffs for low and high confidence, respectively (1). A rating of 50% or greater was considered a positive reading with low confidence for prediction of an aneurysm, whereas a rating of 75% or greater was considered a positive reading with high confidence. With a cutoff of 50%, the number of false-negative cases can be reduced, and the number of false-positive cases may increase. With a cutoff of 75%, the number of false-positive cases decreases, and the number of false-negative cases may increase. To assess the interobserver variability in the interpretation of images, 
statistics were used to measure the degree of agreement between two observers about the presence of an aneurysm. The values greater than 0 were considered to indicate positive agreement; less than 0.4, positive but poor agreement; 0.41–0.75, good agreement; and greater than 0.75, excellent agreement (18).
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Surgery was performed to treat 20 aneurysms in 15 of the 35 patients. In these 15 patients, CT angiographic and surgical findings were compared. Surgery was not performed in patients who were at high risk or refused surgery or who had very small or surgically inaccessible aneurysms.
When the 47 aneurysms were considered individually, the sensitivity of CT angiography for the four observers ranged from 70% to 85% (mean, 80%). Sensitivity levels for middle cerebral artery aneurysms (range, 85%–92%; mean, 90%) were higher than those for others, and sensitivity levels for vertebrobasilar artery aneurysms (range, 50%–83%; mean, 67%) were relatively poor (Figs 1–3). Mean sensitivity levels were 64% for very small aneurysms, 83% for small aneurysms, 95% for medium aneurysms, and 100% for large aneurysms. The highest sensitivity levels among the four observers were 71% for very small aneurysms, 94% for small aneurysms, 100% for medium aneurysms, and 100% for large aneurysms. Even when the cutoff was set at 75% (high level of confidence), the sensitivity of CT angiography remained relatively high (Table 1).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
When detection of at least one aneurysm was assumed positive for the 35 patients (ie, the diagnosis of any one aneurysm would lead to conventional angiography and the detection of any additional aneurysms), the sensitivity ranged from 83% to 91% (mean, 89%) (Table 4). The specificity ranged from 79% to 93% (mean, 88%). The area under the ROC curve (Az index) was 0.93 for observer A, 0.93 for observer B, 0.86 for observer C, and 0.89 for observer D. The ROC curves are shown in Figure 6. The values for interrater variability among four observers indicated excellent agreement (observers A vs B, 0.94; A vs C, 0.81; A vs D, 0.89; B vs C, 0.82; B vs D, 0.91; C vs D, 0.86).
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The use of CT angiography in the detection of aneurysms has been described in several articles, but, to our knowledge, few blinded-reader studies have been published. Schwartz et al (2) evaluated detection of intracranial aneurysms in 21 patients at CT angiography with MIP techniques. They compared the results with those at conventional and MR angiography. They found that all aneurysms 3 mm or larger were depicted at CT and MR angiography; no aneurysms smaller than 3 mm were depicted with either modality. In the prospective study by Alberico et al (7), MIP angiograms depicted 23 of 24 aneurysms in 68 patients, although their technique did not cover the posterior fossa, and mean aneurysm size was as large as 8 mm (in our study, mean size was approximately 5 mm). More recently, Ogawa et al (4) studied 50 patients with 73 aneurysms by using shaded surface display CT angiograms. The authors reported a sensitivity of 24% for the detection of aneurysms smaller than 5 mm. In our series, the highest sensitivity levels among the four observers were 71% for aneurysms smaller than 3 mm and 94% for aneurysms 3–5 mm in diameter; our results were superior to those in the previous studies. We believe there are several advantages with our method. Image quality was better with the volume-rendering technique for 3D reformation, scanning was performed with very thin sections (0.8–1.0 mm), and triggering was optimized to acquire early arterial phase images. These advantages probably contributed to the better detectability rate for aneurysms in our study. The bone editing and isolating of vessels during reconstruction also contribute to better diagnostic performance with CT angiography. In particular, the number of false-negative cases might be markedly reduced.
We used a continuous confidence-judgment scale instead of a discrete scale (1). Although in principle ROC studies can be conducted with either discrete or continuous scales, almost all ROC studies of medical imaging methods have involved data collected with use of a discrete confidence-judgment scale. Rockette et al (14) recommend routine use of continuous scales in radiologic ROC studies because of their potential advantages. The use of continuous scales may be particularly important when the type of abnormality in question is likely to yield highly polarized confidence judgments (19). We chose two different scores from the continuous confidence-judgment scale as the threshold for a positive study. We realized that the observers detected the aneurysms with much higher levels of confidence at CT angiography than they did at MR angiography. In our previous blinded study with MR angiography (1), sensitivity levels decreased by 20%–30% when the cutoff was set at 75% (high confidence) compared to those with a 50% cutoff. In this study with CT angiography, however, differences in the sensitivity levels were less than 10% between the two cutoffs.
Because of current advances in noninvasive imaging, including MR and CT angiography, more unruptured incidental intracranial aneurysms may be discovered. In addition, these relatively new techniques may be used as tools to study the natural course and true frequency in a living sample representative of the total population with the disease (1). Further refining of groups at risk for developing aneurysms may be possible on the basis of findings in a prospective study of the concomitant risk factors or predictive indexes that may affect the frequency of aneurysms and the potential for occurrence of subarachnoid hemorrhage in patients with unruptured aneurysms (1). However, use of CT angiography may be dangerous in a population with a frequency of aneurysms that is much less than that of our study population. If frequency is low, the positive predictive value decreases, and the number of false-positive cases increases. This has enormous implications regarding use of equipment resources and causes patients without disease to undergo invasive imaging, such as conventional angiography, with risk of complications.
Disadvantages with CT angiography, particularly in comparison with MR angiography, include the need for iodinated contrast material and ionizing radiation. The amount of radiation at CT angiography is certainly greater than that at conventional CT but is significantly less than that at digital subtraction angiography (20). The amount of ionizing radiation may not be an important concern in the predominantly older patient population. The administration of iodinated agents with relatively large bolus technique in any setting is potentially problematic (21). Iodinated contrast agents must be used with caution in patients with serious risk factors, such as renal insufficiency, congestive heart failure, or hypersensitivity to contrast material. Timeliness is also important, particularly in patients with subarachnoid hemorrhage. The reconstruction time necessary to depict morphology at CT angiography may be problematic, although it can be reduced with use of newer workstations and experienced technologists. Finally, aneurysms at the skull that arise from the intracavernous or supraclinoid carotid artery may be obscured by bone, calcium, or venous blood, although many aneurysms in this location are not usually considered amenable to resection.
There are two limitations in this study. First, since there were patients who underwent CT angiography but not conventional angiography during the time this study was performed, there is verification bias, and our estimates of diagnostic accuracy are likely inflated. Verification bias is one of the most common biases present in studies of diagnostic accuracy. Second, we used the aneurysm as the unit of analysis to calculate sensitivity. This may be somewhat misleading to clinicians, since sensitivity is usually defined as the proportion of the number of patients with a positive test divided by the number of patients with the disease as derived from a standard two-by-two table. Although the distinction may be subtle, there are potentially important implications for the interpretation of the results by clinicians.
In conclusion, we describe the diagnostic accuracy of 3D CT angiography in the detection of small intracranial aneurysms. In previous reports, CT angiography could not demonstrate aneurysms smaller than 3 mm. With advances in helical CT technology and modification of reformatting techniques, CT angiography can now depict these very small aneurysms. CT angiography is a sensitive technique for detection of intracranial aneurysms 3 mm or larger. Even for aneurysms smaller than 3 mm, CT angiography had relatively good sensitivity. CT angiography has a role as a noninvasive technique for detection of asymptomatic unruptured or ruptured aneurysms.
|
Footnotes |
|---|
Author contributions: Guarantors of integrity of entire study, Y.K., M.T.; study concepts, Y.K., M.T., K.K.; study design, Y.K.; definition of intellectual content, Y.K., M.T.; literature research, Y.K.; clinical studies, K.K., Y.O., K.H., M.O., H.U., T.A., S.I.; data acquisition and analysis, Y.K., K.K., Y.O.; statistical analysis, Y.K.; manuscript preparation and editing, Y.K.; manuscript review, Y.K., M.T.
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. Anzalone, F. Scomazzoni, M. Cirillo, C. Righi, F. Simionato, M. Cadioli, A. Iadanza, M.A. Kirchin, and G. Scotti Follow-Up of Coiled Cerebral Aneurysms at 3T: Comparison of 3D Time-of-Flight MR Angiography and Contrast-Enhanced MR Angiography AJNR Am. J. Neuroradiol., September 1, 2008; 29(8): 1530 - 1536. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A.M. McKinney, C.S. Palmer, C.L. Truwit, A. Karagulle, and M. Teksam Detection of Aneurysms by 64-Section Multidetector CT Angiography in Patients Acutely Suspected of Having an Intracranial Aneurysm and Comparison with Digital Subtraction and 3D Rotational Angiography AJNR Am. J. Neuroradiol., March 1, 2008; 29(3): 594 - 602. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Toyota, K. Iwaisako, H. Takimoto, and T. Yoshimine Intravenous 3D Digital Subtraction Angiography in the Diagnosis of Unruptured Intracranial Aneurysms AJNR Am. J. Neuroradiol., January 1, 2008; 29(1): 107 - 109. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Romijn, H.A.F. G. van Andel, M.A. van Walderveen, M.E. Sprengers, J.C. van Rijn, W.J. van Rooij, H.W. Venema, C.A. Grimbergen, G.J. den Heeten, and C.B. Majoie Diagnostic Accuracy of CT Angiography with Matched Mask Bone Elimination for Detection of Intracranial Aneurysms: Comparison with Digital Subtraction Angiography and 3D Rotational Angiography AJNR Am. J. Neuroradiol., January 1, 2008; 29(1): 134 - 139. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Wada, R. I. Aviv, A. J. Fox, D. J. Sahlas, D. J. Gladstone, G. Tomlinson, and S. P. Symons CT Angiography "Spot Sign" Predicts Hematoma Expansion in Acute Intracerebral Hemorrhage Stroke, April 1, 2007; 38(4): 1257 - 1262. [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]
|
|
|
|
|
|
S. Sakamoto, Y. Kiura, M. Shibukawa, S. Ohba, K. Arita, and K. Kurisu Subtracted 3D CT Angiography for Evaluation of Internal Carotid Artery Aneurysms: Comparison with Conventional Digital Subtraction Angiography AJNR Am. J. Neuroradiol., June 1, 2006; 27(6): 1332 - 1337. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. T. Walker, J. Tsai, T. Parish, B. Tzung, A. Shaibani, E. Krupinski, and E. J. Russell MR Angiographic Evaluation of Platinum Coil Packs at 1.5T and 3T: An In Vitro Assessment of Artifact Production: Technical Note AJNR Am. J. Neuroradiol., April 1, 2005; 26(4): 848 - 853. [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. 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]
|
|
|
|
 |
|
 B. F. Tomandl, N. C. Kostner, M. Schempershofe, W. J. Huk, C. Strauss, L. Anker, and P. Hastreiter CT Angiography of Intracranial Aneurysms: A Focus on Postprocessing RadioGraphics, May 1, 2004; 24(3): 637 - 655. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. V. Jayaraman, W. W. Mayo-Smith, G. A. Tung, R. A. Haas, J. M. Rogg, N. R. Mehta, and C. E. Doberstein Detection of Intracranial Aneurysms: Multi-Detector Row CT Angiography Compared with DSA Radiology, February 1, 2004; 230(2): 510 - 518. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Satoh, K. Onoda, and S. Tsuchimoto Intraoperative Evaluation of Aneurysmal Architecture: Comparative Study with Transluminal Images of 3D MR and CT Angiograms AJNR Am. J. Neuroradiol., November 1, 2003; 24(10): 1975 - 1981. [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 Ane |
