| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiological Sciences, David Geffen School of Medicine at the
University of California, Los Angeles, 10833 Le Conte Ave., Rm. B3-115 CHS,
Los Angeles, CA 90095-1721.
2 Present address: Department of Radiology, University of New Mexico,
Albuquerque, NM 87131.
Received December 14, 2005;
accepted after revision February 28, 2006.
Address correspondence to J. P. Villablanca.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. A blinded retrospective analysis of the presence and size of nine small cerebral arteries on both 2D and 3D CT angiography (CTA) was conducted with 27 patients who had normal findings at CTA and digital subtraction angiography. Scans of 455 arterial segments obtained with either 4-MDCT or 16-MDCT were examined by two independent blinded reviewers. The sensitivity and specificity of CTA for each vessel were established. A chi-square test was used to determine interoperator reliability.
RESULTS. The smallest arterial size reliably detected with MDCT angiography with our imaging and postprocessing protocol was 0.7 mm versus 0.4 mm for digital subtraction angiography. Interoperator reliability for vessel identification with MDCT angiography was 97% without significant differences in detection rates between 4-MDCT and 16-MDCT. Two-dimensional CTA depicted more than 90% of arteries studied but only 63% of anterior choroidal arteries and 27% of recurrent arteries of Heubner. There were no significant differences in mean arterial sizes measured with 2D CTA versus digital subtraction angiography for six of nine arteries. In six of nine arterial segments with a mean diameter of 1 mm or less, fewer arterial segments were visualized on 3D CTA than on 2D CTA.
CONCLUSION. Except for the recurrent artery of Heubner and the anterior choroidal artery, MDCT angiography depicted 90% or more of all examined small intracranial arteries detected with digital subtraction angiography. The mean sensitivity was 0.91, and the mean specificity was 0.7.
Keywords: catheter angiography • cerebral arteries • CT angiography • 2D • 3D • digital subtraction angiography
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In a retrospective blinded MDCT angiographic analysis of 672 intracranial arteries, Bash et al. [1] found a sensitivity and positive predictive value of 98% and 100% for intracranial stenosis compared with catheter angiography (digital subtraction angiography) and MR angiography (3D time-of-flight MR angiography). These investigators, however, excluded from evaluation any arteries that may have been smaller than 1 mm, because such vessels were not consistently seen with all three imaging techniques evaluated. Using CTA and digital subtraction angiography, Skutta et al. [4] retrospectively examined 2,205 arterial segments in patients with suspected atherosclerosis and concluded that CTA can reliably depict vessels as small as 0.7 mm in diameter if targeted maximum-intensity-projection images are used and further enhanced by the use of source images to detect these smaller arteries. Those investigators reported 80% identification of the ACoA, 76% identification of the PCoA, 97% identification of the SCA, 89% identification of the PICA, and 63% of the AICA with the combined approach. They did not report on the ability of CTA to depict the anterior choroidal artery, ophthalmic artery, or recurrent artery of Heubner. It is known that all of these vessels can participate in pathologic processes leading to stenosis and occlusion with resultant stroke. These processes include atherosclerosis, embolism, vasculitis, and vasospasm.
Our purpose was to determine the average size, size range, and rate of detection of small intracranial arteries with both MDCT angiography and digital subtraction angiography using commonly available MDCT technology and scanning protocols. Our goal was to achieve a better understanding of the capabilities and limitations of both MDCT angiography and digital subtraction angiography in the evaluation of small intracranial arteries. To our knowledge there are no current reports on the frequency with which MDCT angiography can depict small normal intracranial arteries or on the accuracy and reproducibility in measurement of arterial diameters of MDCT angiography compared with digital subtraction angiography.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CTA
Helical scans were obtained with 4-MDCT or 16-MDCT. The 4-MDCT (LightSpeed,
GE Healthcare) imaging protocol was pitch, 3.75; slice collimation, 1.25 mm;
reconstruction interval, 0.625 mm; field of view, 18 cm; matrix size, 512
x 512; 120 kVp; 300 mA. The 16-MDCT (Sensation 16, Siemens Medical
Solutions) protocol was slice collimation, 1 mm; feed rate, 6.8 mm per
rotation; reconstruction interval, 0.5 mm; field of view, 18 cm; matrix size,
512 x 512; effective tube current, 320 mAs at 120 kVp. With this
protocol, average voxel size was x-axis, 0.35 mm; y-axis,
0.35 mm; z-axis, 0.625 mm. All studies were performed with the
SmartPrep option (GE Healthcare) for automatic bolus arrival detection or with
manual calculation of antecubital-to-midcervical arrival time of a test
injection of contrast material. The test injection was a 15-mL bolus of
iohexol (Omnipaque 350, Nycomed) into the right antecubital vein through a
22-gauge angiocatheter at a rate of 3 mL/s, upfront delay of 10 seconds, 15
images for a 1-second scan, 1-second interscan delay, and scanning at midneck
at 80 kVp and 80 mA. Timed injections allow helical scanning to coincide with
peak intracerebral arterial contrast opacification. The angiographic portion
of the study was performed with an IV contrast dose delivered through a
22-gauge angiocatheter into the right antecubital fossa. Iohexol was injected
at a rate of 3 mL/s for a total volume of 3 mL/s multiplied by total exposure
time, generally approximately 95-125 mL total volume for 4-MDCT and 60 mL
total volume for 16-MDCT.
Digital Subtraction Angiography
Digital subtraction angiography was performed with standard access
techniques, angiographic runs being performed for the bilateral internal
carotid and bilateral vertebral arteries when accessible. Oblique projections
were obtained as needed to clarify angiographic anatomy. Cross-compression
runs were performed whenever necessary for visualization of the ACoA. Oblique
and magnification views were obtained as needed to clarify the anatomic
features of the vessels. The IV contrast medium was iohexol (Omnipaque 300)
injected at 80% concentration and 7-10 mL total volume with manual injection
rates of 4-5 mL/s for both the anterior and posterior circulation
injections.
CTA and Digital Subtraction Angiography Image Analysis
A retrospective blinded quantitative analysis of data acquired on MDCT
angiography and digital subtraction angiography of the brain was performed
with identification and measurement of all listed arteries for each patient. A
total of 455 arterial segments were examined. For CTA data, each artery was
evaluated as visualized or not visualized on both the gray-scale 2D source
images and 3D volume-rendered images. Quantitation was performed with only the
2D gray-scale multiplanar reformatted images in a plane perpendicular to the
long axis of the vessel. Measurements were made with the internal digital
caliper of the workstation (Vitrea 2.7, Vital Images) with a
manufacturer-specified accuracy for measurements of ± 10% for
measurements less than 2 mm, ± 5% for measurements of 2-10 mm, and
± 2% for measurements greater than 10 mm. All measurements were made by
visual estimation of full-width at half-maximum of the slice-sensitivity
profile of the vessel borders at an approximate window setting of 1,000 and a
level setting of 500.
Digital subtraction angiographic analysis consisted of visual determination of vessel presence and, if the vessel was visible, quantitative analysis of arterial diameter. Because conventional angiograms do not contain inherent image scales, direct measurements were not possible. The method of including an object of known size in the field of view, although appropriate for most clinical applications, was not considered optimal for this study because of difficulty in properly correcting for the phenomenon of differential magnification between the object and the target vessel. Therefore measurements were made with an equation in which an average standard diameter of the supraclinoid segment of the internal carotid and basilar arteries is assumed. These values were obtained from previously published values based on results of histopathologic studies [5]. The diameter of the reference artery was measured from the study film with a digital caliper. A proportionality equation was set up to solve for the diameter of the selected artery when the measured diameter of the artery was known from the conventional angiogram and the calculated ratio of the supraclinoid internal carotid artery measurements.
All MDCT angiographic measurements were made by two independent reviewers blinded to measurements made by each other and to digital subtraction angiographic measurements. Digital subtraction angiographic measurements were made by one reviewer blinded to all MDCT angiographic data and measurements. MDCT angiographic and digital subtraction angiographic image data were presented in random order with patient identifiers removed.
The following factors were tabulated for all patients and all arterial segments: ability to visualize arterial segment on 2D and 3D CTA, diameter of arterial segment on MDCT angiographic 2D gray-scale imaging, ability to visualize an artery with digital subtraction angiography, diameter of arterial segment on digital subtraction angiography, and image quality on 2D and 3D CTA and digital subtraction angiography. The nine arteries selected for analysis were chosen because they were likely to contain arterial segments that were both within and below the expected detection thresholds for MDCT angiography.
MDCT and digital subtraction angiographic studies were rated as excellent if no motion or beam-hardening artifacts were present and if the scan was obtained during peak arterial contrast opacification, yielding homogeneously dense arterial opacification and sharp vessel borders. Good image quality was assigned to images with motion or beam-hardening artifacts or intraluminal contrast inhomogeneity that did not interfere with image analysis. Fair and poor ratings were given to studies with patient motion, beam-hardening artifacts, or bolus-timing problems that precluded evaluation of vessels because of blurred image borders and low or highly heterogeneous intraluminal contrast density.
Analysis
Barnard's exact test and exact confidence intervals were used to evaluate
differences in proportions of vessels and diameters on 2D and 3D CTA and
digital subtraction angiography. Standard contingency tables for the
sensitivity and specificity of MDCT angiography for each vessel were
established for vessel detection rates on MDCT angiography and to obtain 95%
CI estimates. A chi-square test was used to determine interoperator
reliability of MDCT angiographic measurements and detection rates.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The interoperator reliability coefficient for MDCT angiography vessel identification was 97% between reviewers 1 and 2 and 0.95 for 2D CTA diameter measurements compared with digital subtraction angiographic measurements. There was no statistically significant difference in detection rates for any vessel segment in comparisons of 4-MDCT (11/27) and 16-MDCT (16/27) systems with our imaging device and scan protocol combination.
Table 1 lists the arterial segments evaluated, the frequency with which each artery was seen on MDCT angiography (both 2D and 3D) and digital subtraction angiography, and the size range and mean for each arterial segment according to technique. A maximum of 455 vessel segments were evaluated with MDCT angiography and 440 with digital subtraction angiography. For MDCT angiography, motion artifacts precluded evaluation of both recurrent arteries of Heubner in two patients, beam-hardening artifacts obscured the FP arteries in one patient, and two patients had a duplicated ACoA. Depending on the vessels into which injections were made and individual patient anatomic features, not all artery segments were evaluable with digital subtraction angiography in all patients, as shown in Table 1. All measurements were rounded to the nearest tenth of a millimeter. A p value was calculated for the difference between 2D CTA and digital subtraction angiography measurements.
|
As expected, digital subtraction angiography depicted arteries below the detection threshold for MDCT angiography. The smallest arterial diameter reliably depicted with MDCT angiography with our scan and post-processing protocol and the source image data was 0.7 mm, whereas for digital subtraction angiography the smallest artery depicted in this group of arteries was 0.4 mm in diameter. The smallest arteries visible with MDCT angiography were the anterior choroidal artery (Fig. 1A, 1B) and recurrent artery of Heubner, which had average arterial diameters of 0.91 and 0.95 mm on CTA versus 0.88 and 0.77 mm on digital subtraction angiography.
|
|
|
|
|
|
MDCT angiographic visualization of arterial segments was not equal across the arteries examined. Three-dimensional CTA visualization was inferior in relation to 2D CTA source image visualization for the FP, PCoA, anterior choroidal artery (Fig. 3A, 3B, 3C), and AICA. These arteries had a combined average size of 0.99 mm compared with a combined average size of 1.1 mm for the other arteries. In six of nine arterial segments, an average of 4.5 fewer arterial segments across the group of patients were visualized on 3D CTA compared with 2D CTA. This finding occurred for arteries with smaller size ranges and with a mean arterial diameter of 1 mm or less. Three-dimensional CTA was equal to 2D CTA source imaging in identification of the ophthalmic artery, ACoA, recurrent artery of Heubner, SCA, and PICA, which had a combined average size of 1.1 mm.
|
|
|
Table 3 shows that the sensitivity of MDCT angiography in detection of the FP artery is comparable for 2D and 3D image review formats, although both image review methods are associated with a high false-positive fraction. This finding suggests that the FP artery can be difficult to identify definitively on both 2D and 3D images compared with larger, more readily identifiable arteries, such as the ophthalmic artery, which has a more predictable course and location and, in some instances, a more reliable relation to specific bone structures.
|
Although the sensitivity of 2D CTA was higher than that of 3D CTA for identification of the PCoA (87% vs 78%), the difference was not statistically significant (p = 0.25). The specificity was twice as high for 3D CTA as for 2D CTA for the PCoA, reflecting the inherent advantage of 3D images compared with 2D source data in rapid visualization of the origin and insertion point of arterial segments. This difference also did not reach statistical significance (p = 0.19).
The specificity of both 2D CTA and 3D CTA in detection of the anterior choroidal artery varied from 50% to 70% with a relatively high false-negative fraction. This finding likely occurred because this arterial segment was among the smallest selected for measurement, nearly one third of the arterial segments having a diameter less than the detection threshold of CTA. The reference standard, digital subtraction angiography, depicted 88% of all possible anterior choroidal arteries. In all, 2D CTA depicted 63% of all anterior choroidal artery segments seen with digital subtraction angiography (Table 2). In contrast to digital subtraction angiography, 3D CTA depicted less than one half (42%) of these segments. The lower detection rate of 3D digital subtraction angiography was likely related to difficulties encountered by volume-rendering programs in segmenting very small arterial segments with meandering courses across multiple slices.
The high sensitivity and specificity of both 2D and 3D CTA in detection of the ophthalmic artery (Table 3) was likely related to the comparatively larger size of this artery, averaging a mean diameter of 1.1 mm, compared with very small arteries, such as the anterior choroidal artery and recurrent artery of Heubner, which on digital subtraction angiography and 2D CTA averaged 0.9 and 0.8 mm, respectively (Table 1). Although the sensitivity of both 2D and 3D CTA for the ACoA were 100%, the specificity was negatively affected by the fact that the reference standard, digital subtraction angiography, depicted only 65% of all possible ACoAs (Table 2). This finding was in contrast to those for both 2D and 3D CTA, which depicted 97% of possible arterial segments (Fig. 4A, 4B), approximately 140% of the number depicted with digital subtraction angiography. This finding explains the large false-positive fraction for MDCT angiography and the artificially low apparent specificity.
|
|
The sensitivity of both 2D and 3D CTA for the PICA was 84%. The specificity was low for both 2D and 3D CTA (Table 3). This finding might not have been an accurate reflection of the ability of CTA to reliably depict this arterial segment because the true-negative fraction was only 2/47.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We postulate that the higher detection rates derive from differences in scan protocol and improvements in postprocessing software and hardware. Image acquisition parameters play an important role in measurement error and vessel conspicuity. Suzuki et al. [6] found that absolute measurement errors were lowest with higher densities of intravascular contrast medium and that overall measurement errors were minimized when a soft convolution kernel was used for image reconstruction independent of the density of the contrast material within the model. In a subsequent article, Suzuki et al. [7] showed that measurement errors were minimized when smaller sizes of display field of view were used. Other investigators [8] have found that overlapping reconstructed slices decrease stairstep artifacts in 3D reconstructions. In a phantom study, Kuszyk et al. [8] found that the most accurate representations of true phantom diameters were obtained with an intravascular attenuation value of 325 H. Takhtani [9] emphasized that although introducing reconstruction overlap does not provide a smaller slice thickness than that provided by the detector collimator, using 50% overlapping reconstructed slices does serve to minimize stairstep artifacts in 3D images. Other authors [10] have emphasized the importance of a timing injection because delay of arrival of contrast medium from the antecubital fossa to the midneck can vary from 12 to 54 seconds in healthy subjects. The importance of optimal contrast density was emphasized by Claves et al. [11], who evaluated carotid arteries and found an optimal intraluminal contrast density for both ionic and nonionic contrast media.
Although depiction of small intracerebral arteries was generally as good with MDCT angiography as with digital subtraction angiography, the more distal vessel segments at any vascular diameter evaluated were uniformly less well seen on MDCT angiography than on digital subtraction angiography. This finding was likely due to the lower spatial resolution of current MDCT angiographic techniques compared with digital subtraction angiography. For this reason, we do not advocate the use of MDCT angiography in the initial diagnosis of neurovascular abnormalities that can manifest themselves in the distal circulation, such as cerebral vasculitis, amyloid angiopathy, and mycotic aneurysms. Although we routinely receive requests to perform brain MDCT angiography for excluding cerebral vasculitis, we decline these requests, instead advising the clinician to perform definitive catheter angiography.
For the FP artery, PCoA, anterior choroidal artery, ophthalmic artery, and recurrent artery of Heubner segments in this study population, there were no statistically significant differences in mean arterial sizes measured with 2D CTA as opposed to digital subtraction angiography. This finding suggests the phenomenon of artifactual luminal eccentricity was likely not a significant factor in our analysis. This phenomenon has been reported to affect stenosis calculations with MDCT angiography when the vessel segment is perpendicular to the z-axis of the scanner [12]. In addition, our measured arterial diameters for both MDCT angiography and derived arterial diameters for digital subtraction angiography correlated fairly well with published values. For instance, using 71 white orbits, Lang and Kageyama [13] found a mean sex-averaged diameter of the ophthalmic artery measured 2 mm from the origin to be 1.4 ± 0.5 mm, compared with a somewhat lower 1.2 ± 0.3 mm for digital subtraction angiography and 1.1 ± 0.1 mm for MDCT angiography in our study. This difference may be explained by the fact that because both MDCT angiography and digital subtraction angiography are vessel cast techniques, internal diameter is measured, whereas in pathologic specimens, external diameter is measured, adding arterial wall thickness to each measurement. In agreement with others [14], we found the PCoA to be a single vessel in every patient in whom that arterial segment was present.
MDCT angiography depicted 63% of anterior choroidal arteries that were depicted by digital subtraction angiography, indicating MDCT angiography can be used to interrogate this arterial segment with a detection rate nearly two thirds that of catheter angiography. The raw rate of detection of the anterior choroidal artery with 2D CTA was 56% with a sensitivity of 53%. In contrast, digital subtraction angiography depicted 89% of anterior choroidal arteries. CTA performed poorly in evaluation of the recurrent artery of Heubner, depicting only 27% of arteries found with digital subtraction angiography with a 22% sensitivity and 93% specificity of 2D CTA. Both the anterior choroidal artery and recurrent artery of Heubner are very small vessels, averaging 0.9 mm on both MDCT angiography and digital subtraction angiography, the size range being 0.4-1.5 mm on digital subtraction angiography. In a cadaveric study of 25 brains, Rhoton et al. [15] found the mean diameter of the anterior choroidal artery was 1.2 mm with a range of 0.7-2.0 mm, in good agreement with our ranges. All of the anterior choroidal arteries visualized with digital subtraction angiography but not with MDCT angiography were smaller than 0.8 mm. The smallest anterior choroidal artery visualized with MDCT angiography was 0.8 mm, whereas for digital subtraction angiography that value was 0.4 mm. These results indicate that with the protocol studied, MDCT angiography should not be used to interrogate suspected pathologic conditions involving any of the perforator arteries arising from the A1, M1, PCoA, or P1 segments. These arterial segments are generally smaller than the current detection threshold of MDCT angiography. In addition, lack of visualization of the anterior choroidal artery or recurrent artery of Heubner with MDCT angiography should not be construed to indicate evidence of occlusion. In addition, both 2D and 3D CTA generally depicted only the proximal 1-2 cm of the smallest arterial segments, compared with depiction of extended segments with digital subtraction angiography. This finding indicates that although MDCT angiography can be used to document the patency of these small arteries, the technique is usually not reliable for determination of the presence of stenosis beyond the first centimeter of a vessel.
There were no statistically significant differences in mean arterial sizes measured with 2D CTA versus digital subtraction angiography except for the SCA, AICA, and PICA (p < 0.001). In these vessels the upper ranges of measured diameters for digital subtraction angiography were consistently larger than for 2D CTA. We speculate this difference might have been due to a systematic measurement error introduced by use of a reference artery diameter for the basilar artery that was not as representative of true arterial diameter as that used for the internal carotid artery [5].
Compared with digital subtraction angiography, MDCT angiography depicted more anterior to posterior and left to right collateral arterial segments, such as the ACoA and PCoA (Table 2). This difference was most pronounced for the ACoA, which was well visualized in 97% of patients, compared with 65% for digital subtraction angiography. We suspect the difference was caused by the difficulty in obtaining satisfactory opacification of arterial segments with balanced flow through communicator vessels during catheter angiography. Visualization of these arterial segments is typically performed with the cross-compression technique, which for various reasons can be accomplished with variable success. Because CTA is a vessel cast technique, all arterial segments are filled simultaneously, and therefore central collateral arteries (ACoA, PCoA) are excellently opacified. We believe the ability of MDCT angiography to easily and consistently depict communicator arteries of the circle of Willis is an important advantage over digital subtraction angiography.
Differences in visualization rates observed between 2D and 3D CTA images is likely related to the limitations of current volume-rendering programs in accurately depicting the longitudinal course of very slender arterial segments as they course through adjacent slices in a volume. The result is volume-averaging artifacts and decreased contrast density within very slender arteries. Therefore we recommend using 2D source images in addition to a surface- or volume-rendering approach whenever it is desirable to identify an arterial segment smaller than 1.5 mm in diameter. We chose volume rendering as the 3D CTA visualization method over surface-shaded display and maximum intensity projection because others [16] have shown that volume-rendering methods are more accurate for stenosis measurements in vessels 2-4 mm in diameter and are superior in measurement of arteries 0.5-1.5 mm in diameter. Addis et al. [16] and Lo et al. [17] found the measurement error of 3D volume rendering to be low, 0-2.5%. We have found measurements obtained from 3D images may be affected by artifacts arising from sub-optimal window and level settings and therefore prefer to rely on the 2D source or multi-planar reformatted gray-scale images for all measurements. These artifacts include pseudostenosis and vascular blooming, apparent fusion of two closely spaced vessels, and disappearance of smaller arteries.
Compared with the findings of Skutta et al. [4], our findings were higher identification rates for the ACoA, PCoA, SCA, and AICA on 2D CTA source images. These findings are likely related to use of higher contrast concentration and contrast injection rates. In the study by Skutta et al., no timing injection was used, possibly leading to scanning during a non-peak contrast opacification phase. Furthermore, use of low tube current may have led to increased image noise and decreased vessel conspicuity. Pixel size cannot be calculated but if greater than necessary may also have negatively affected spatial resolution. It is clear that compared with digital subtraction angiography, MDCT angiography does not depict several very small arteries of potential importance. MDCT angiography, however, does have several advantages over 2D digital subtraction angiography, including lower cost, a better safety profile, and an infinite number of viewing angles compared with 2D digital subtraction angiography. Some arteries also are too small to visualize with current conventional catheter angiography techniques [18].
Except for the recurrent artery of Heubner and the anterior choroidal artery, MDCT angiography depicted 90% or more of all examined small intracranial arteries detected with digital subtraction angiography, with a mean sensitivity of 91% and a mean specificity of 70%. Two-dimensional CTA has a higher sensitivity than 3D CTA in the detection of small brain arteries, but 3D CTA has a higher specificity than 2D CTA and can aid in identification of small arteries in the brain.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
