Purpose. The purpose of this study was to develop a valid, reliable and accurate system of measurement of abdominal aortic aneurysms, using volumetric analysis of x-ray computed tomographic data. Materials and Methods. This study evaluates illustrative cases, and compares measurements of AAA phantoms, using standard two- dimensional versus volumetric methods. To validate the volumetric analysis, four phantom aneurysms were constructed in a range of diameters (4.5 - 7.0 cm) which presents the greatest management challenge to the clinician. These phantoms were imaged using a Toshiba Xpress SX helical CT. Separate scans were obtained at conventional (10 mm x 10 mm) and thin slice (5 mm x 5 mm) collimations. The thin slices were reconstructed at 2 mm intervals. Data from each of the 96 scans were interpreted using a standard two-dimensional approach, then analyzed using task- oriented volumetric software. We evaluate patient assessments, and compare greatest outer diameters of phantoms, by standard versus volumetric methods. Results. Qualitative differences between solutions based on standard versus volumetric analysis of illustrative patient cases are substantial. Expert radiologists' standard measurements of phantom aneurysms are highly reliable (r2 = 0.901 - 0.958; p less than 0.001), but biased toward significant overestimation of aneurysm diameters in the range of clinical interest. For the same phantoms, volumetric analysis was both more reliable (r2 = 0.986 - 0.996; p less than 0.001), and more accurate, with no significant bias in the range of interest. Conclusions. Volumetric analysis promotes selection of more valid management strategies, by providing vital information not otherwise available, and allowing more reliable and accurate assessment of abdominal aortic aneurysms. It is particularly valuable in the presence of aortic tortuosity, vessel eccentricity, and uncertain involvement of critical vessels.
The result of ruptured aortic aneurysm is patient death 70-90% of the time. 1,2 Many factors (eg. alterations in volume fractions of collagen and elastin, influence of reflected pressure waves, diffusion of nutrients across mural thrombus) play a role in the natural history of abdominal aortic aneurysms, 3-8 We have not yet found a way to monitor these factors and reliably predict their failure. The decision to repair an abdominal aortic aneurysm (AAA) is therefore based primarily on diameter; the only parameter so far identified which is significantly related to rate of aneurysmal expansion and rupture. 9-11 An AAA expanding faster than 5 mm in a six month period is perceived to be at higher risk for imminent rupture, usually prompting surgical repair. 12 Cross-clamping above the renal arteries, required for aneurysms which involve renal or visceral branch vessels, is associated with increased morbidity and mortality from increased cardiac stress, decreased renal perfusion, and increased technical difficulty.13 Measurement of abdominal aortic aneurysms (AAA), and evaluation of branch vessel involvement, are conventionally derived from two- dimensional transverse computed tomographic images, although many centers still rely predominately upon ultrasonic evaluation for screening and follow-up of asymptomatic patients. 14 Measurements discriminate patients who receive surgical or radiologic intervention from those managed conservatively. Standard approaches vary most in measurement, or in assessment of branch vessel involvement, where aortic tortuousity and eccentricity are not adequately recognized.
The patient with an asymptomatic AAA who survives the surgical procedure has a significantly improved life expectancy approaching age-related norms, 15,16 although death from myocardial infarction still limits expected five-year survival.17 The objective of effective therapeutic intervention is to treat an asymptomatic AAA when the risks associated with treatment are less than the risks of further conservative management. The risks of treatment are fairly well-known, with mortality related to elective repair of 2.6-5.5% in larger centers. 17,18 Accurate prediction of the probability of a catastrophic rupture, then, is central to effective management.18,19 Determination of this probability depends upon a valid, reliable and accurate system of measurement of AAA size. Assessment of visceral vessel involvement is also critical in preinterventional planning. Evolving endovascular approaches to AAA repair require even more detailed and complex understanding of AAA morphology.20,21
Volumetric approaches to the evaluation of organ structure and function have been extensively explored using Mayo's Dynamic Spatial Reconstructor 22 and electron- beam CT.23 With the recent commercialization of slip-ring technology (e.g. helical CT), methods for visualization and quantitation can be widely applied to clinically relevant questions. It is the premise of this work that task-oriented use of subsets of volumetric tools can bring important new information to the evaluation of patients with suspected aortic aneurysms. While three-dimensional representations, such as shaded surface displays and maximum intensity projections, have been widely used in association with volumetric evaluations, we believe that their overuse may obscure the true value of volumetric imaging. Volumetric analysis interrogates both visually and quantitatively a digital representation of the body region of interest acquired in stacked cross-sectional images.
In response to contemporary patient problems recently encountered in our medical center, we have evaluated four illustrative cases, employing three volumetric methodologies: oblique sectioning and region of interest analysis; geometric analysis of outer aortic boundaries and vessel tortuosity; and volumetric display. These cases, presented below, illustrate the magnitude of error inherent in standard analysis of two-dimensional transverse computed tomographic images, as well as the potential adverse impact of ambiguous interpretation of qualitative information vital to appropriate management. Measurements employing volumetric methodologies have been extensively validated in other applications.24 As a first step in the direct validation of volumetric analysis of abdominal aortic aneurysms, we present a study validating volumetric analysis in phantom tubular structures with complex three- dimensional morphology.
Case 1: Diagnostic Dilemma. Patient 1 is a 77 year old female who had repair of an infrarenal abdominal aortic aneurysm with a dacron tube graft seven years previously. She now presents with a palpable abdominal mass. From two-dimensional transverse computed tomographic images (Figure 1), it is unclear whether there is a large pararenal aneurysm, an anastamotic pseudoaneurysm, or simply aortic tortuosity.
To determine the geometry of the pararenal aorta (Figure 2, "A"), planes orthonormal to the true centerline (as in "B") were evaluated. This eliminated artifactual overestimation of aneurysmal diameter due to extreme tortuosity, and ruled out pseudoaneurysm. The true diameter of the aneurysm ("C," = 4.8 cm) did not require intervention at this time.
Figure 1. Consecutive 10 mm x 10 mm helical CT sections are obtained through the abdominal aorta at level "A". An infrarenal aneurysm repair has previously been performed.
Figure 2. Volumetric analysis controls for error from vessel tortuosity and eccentricity (see text for figure labels).
Case 2: Intervention or Conservative Management. Patient 2 is a 76 year old male with a large AAA and intraperitoneal spread of colon cancer. The threshold diameter for elective repair, given his underlying condition, should be substantially greater than in an otherwise healthy patient.19 Standard two dimensional analysis (Figure 3) could not determine whether this aneurysm was tortuous, and 6 cm in diameter, or eccentric, and 8 cm. Both extremes have been reported in recent CT examinations of this patient.
Figure 3. Consecutive 5 mm x 2 mm helical CT images are obtained through the abdominal aorta, and the diameter of the aneurysm is measured. The diameter was reported to be 6 x 8 cm, through the long ("A") and short ("B") axes.
Volumetric analysis (Figure 4) of the aneurysm revealed that uncertainties of measurement were due to a combination of tortuosity and eccentricity. The greatest orthonormal diameter (6.7 cm), as determined by volumetric analysis, was below the surgeon's threshold for repair at this point in time.
Figure 4. Vessel tortuosity and tubular eccentricity degrade accuracy of standard two-dimensional measurements. Volumetric analysis allows examination perpendicular to the true centerline of the aneurysm, avoiding such errors.
Figure 5. Consecutive 5 mm x 2 mm helical CT images are obtained through the abdominal aorta.
Case 3: Surgical Approach. Patient 3 is a 76 year old male with a large abdominal aortic aneurysm. Serial CT examinations reported that the aneurysm begins "just below" the renal arteries. From standard transverse images (Figure 5), it is unclear whether there is a sufficient "neck" to clamp the aorta above the aneurysm but below the renal arteries during repair. Suspicion of suprarenal extension requires the surgeon to cross-clamp the aorta above the renal arteries, a technically more demanding procedure with increased operative morbidity and mortality. Incorrect assessment may alter the operative approach and increase the surgical risk for the patient. 25
Figure 6. Shaded surface displays and multiplanar reformatting provided information not available, even in retrospect, from standard transverse two-dimensional images.
Combined information from shaded surface display (SSD) and oblique coronal sections (Figure 6) demonstrated an adequate neck for clamping below the renal arteries. This allows the surgeon an optimal approach, reducing operative time and surgical risk.
Figure 7. Consecutive 3 mm x 3 mm helical CT images are obtained through the abdominal aorta.
Case 4: Standard or Endoluminal Repair. Patient 4 is a 75 year old male with severe atherosclerotic coronary artery and peripheral vascular disease with a large saccular abdominal aortic aneurysm. Because multisystem involvement significantly increases his operative risk, he was considered for endoluminal repair of his aneurysm by percutaneous transluminal placement of a coated stent graft. From standard transverse images (Figure 7), it was unclear whether a distal "neck" sufficient to seat an endoluminal stent graft was present above the aortic bifurcation. Again, incorrect assessment may alter the interventional approach and increase risk for the patient.
Figure 8. Volumetric analysis from reformatted oblique planes. ("A"- inferior limit of the aneurysm; "B" - superior limit of the aortic bifurcation; "C" - proximal common iliac arteries; "D" - shaded surface display of aneurysm and spine.)
Ambiguity regarding aneurysmal involvement of the bifurcation persisted despite shaded surface display rendering. Analysis of reformatted oblique planes through the volume of interest (Figure 8) defined an insufficient distal neck for seating an endoluminal stent graft. The patient was treated by conventional surgical aneurysm replacement.
Standards for measurement of abdominal aortic aneurysms are usually measurements taken at surgery or at autopsy.26 Altered dimensions due to deviation from normal distending pressures in these two situations may be substantial. As the above patient cases illustrate, moderate to severe curvature of the ectatic or aneurysmal aorta may introduce substantial error in measurements taken in an arbitrary plane (such as the transverse plane conventionally used in standard CT imaging). Any diameter on a transverse image that is not exactly parallel to the radius of the curve (or perpendicular to the local long axis) at that point will be inaccurately represented.27 Intraoperative access to relevant angles for measurement perpendicular to the true centerline is severely limited. For these reasons, current "standards" are inaccurate. The aim of this study was to develop a valid, reliable and accurate system of measurement of abdominal aortic aneurysms, using volumetric analysis of x-ray computed tomographic data. For this static validation study, we elected to construct phantom aneurysms for imaging and analysis with stable true diameters perpendicular to the local long axis amenable to direct, accurate measurement.
Phantom aneurysms (Jim's Instrument Manufacturing, Inc., Iowa City, Iowa) were hand-blown from Plexiglas tubing 25 cm in length with an outer diameter of 2.54 cm, and a thickness of 0.32 - 0.64 cm. To introduce inherent tortuosity, the phantoms were allowed to bend during heating in all three spatial planes. To force a controlled degree of tortuosity, the phantoms were imaged according to a 3 x 2 x 2 matrix of rotations in the transverse (0, 45, 90 degrees), coronal (0, 30 degrees), and sagittal (0, 20 degrees) planes. Eccentricity, rather than concentricity, was encouraged in the phantom circumference. The phantoms were filled with sonographic gel to simulate intravascular contrast. For imaging, they were suspended in a plexiglass cylinder filled with potato flakes which served to stabilize the phantom.
The phantom aneurysms (Figure 9) were measured with electronic micrometer calipers. Each was rotated into an oblique plane to maximize the outer diameter perpendicular to the local long axis (greatest orthonormal diameter), and measured by at least four observers. Measurements were then averaged over at least eight observations. The average greatest orthonormal diameters were as follows:
Figure 9. Plexiglas phantom aneurysms.
The phantoms were imaged on a Toshiba Xpress SX helical scanner in helical mode at a pitch of 1:1, using a medium (320 mm) field of view without postprocessing zoom, with kVp of 120 and a tube current of 100 mA. (The only difference from the normal clinical protocol was the use of a lower x-ray tube current.) Serial images were acquired according to the matrix of rotations described above, using both conventional thick-slice (10 mm x 10 mm) and reformatted thin-slice (5 mm x 5 mm reconstructed at each 2 mm) protocols, which resulted in a total of 24 scans of each of the four plexiglass phantoms. All 10 mm x 10 mm images, and every third 5 mm x 2 mm (reconstructed) image, were filmed for two-dimensional analysis, and transferred to a dedicated UNIX workstation for volumetric analysis.
Standard two-dimensional analysis. Two radiologists expert in cardiovascular imaging independently scored the four AAA phantoms. Each of the 96 studies was filmed and placed on a separate alternator view-bank in random order, for scoring. Following the technique of Lederle et al.,28 the maximum external diameter in any dimension was interpolated to the nearest millimeter, using calipers and the reference scale provided with each image, and entered on the score sheet provided. If the scorer determined that eccentricity of a phantom aneurysm was attributable to obliquity through the scanning plane, the greatest diameter was defined as the diameter perpendicular to the direction of tortuosity within the image series, and a notation signifying the variation in scoring was made.
Volumetric analysis. Image data from both the thick- and thin-slice scans was transferred from the CT network (CEMAX) to a UNIX-based workstation for analysis by VIDA 29 (Volumetric Image Display and Analysis), a modular system for multi- dimensional image display and analysis. The IMPROMPTU module 30 was used to develop an algorithm for semi-automated segmentation of the phantom aneurysms. (Outer walls of patient aneurysms are currently segmented by hand.) The segmented image was interpolated to cubic voxels using the Shape-based Interpolation module.
The phantoms were measured using VIDA's Tube Geometry Analysis (TGA) module (Figure 10). TGA was developed to make three dimensional measurements of pre- segmented eccentric tubular structures at any specified spatial position. TGA displays three reference maximum intensity projections (transverse, coronal, and sagittal MIPs) of the pre-segmented vessel. Using one of the three MIP projections, two points are placed within the abdominal aorta at the extremes of the segment of interest to approximately define the vessel's center. The two endpoints are simultaneously adjusted in all three projections such that the points fall within the vessel center in all projections. Based on the user-estimated center, TGA uses the line connecting the two endpoints to calculate a perpendicular plane and find the center of mass of the vessel voxels within that plane. Two line segments are generated from the new centerpoint and the original endpoints. This process is repeated through four to five iterations, to determine 31 to 63 points which closely approximate the vessel's true centerline. Oblique planes perpendicular to the local long axis (or parallel to the radius of curvature) at the centerline points are calculated and used to graphically display the greatest orthonormal diameter of each of the oblique planes. The peak of this graph is the greatest outer wall diameter of the aneurysm.
The radiologists readily identified the apparent greatest diameter of the phantom aneurysms within two-dimensional transverse slices, and even agreed upon the same slice most of the time. However, detection of or agreement upon the degree of tortuosity was much more difficult. The reliability of their measurements is reflected in the high correlations (Table 1). These measurements are biased, however, and tend to significantly overestimate the true values for the size range of aneurysms which most often present a management dilemma for the physician (Figure 11). Expert radiologists' measurement errors are repeatable, not attributable to the bias of a single reader. Volumetric analysis has even greater reliability, reflected as less dispersion of measurements about the regression line (Figure 11). Additionally, regression analysis demonstrates substantially closer agreement with the identity line, with a slope nearer unity and an intercept nearer zero. (Table 1) This translates to a substantial increase in accuracy for volumetric analysis versus the standard two- dimensional transverse computed tomographic system of measurement considered the current "gold standard".
Figure 10. Tube Geometry Analysis. (A. Sagittal maximum intensity projection [MIP], B. Coronal MIP, C. Transverse MIP, D. Greatest orthonormal diameter)
Reliability and accuracy significantly improved when volumetric analysis was applied to the higher resolution, thin-slice images compared to conventional thick- slice images. RadiologistsÕ performance did not change significantly with the additional information. (Table 1)
Table 1. Regression analysis was performed on greatest diameters measured by expert readers (two-dimensional analysis), and by volumetric analysis, versus true diameters of the four aneurysm phantoms (*n = 48).
Figure 11. Regression analysis. Comparison of calculated diameters from volumetric software (VIDA), Reader 1, and Reader 2 versus predicted (true) diameters for four phantom aneurysms using the thin-slice reformatted (5 mm x 2 mm) protocol.
In the current study, we demonstrate the utilization of volumetric image display and analysis tools in the preoperative assessment of abdominal aortic aneurysms. We have shown that qualitative information vital to appropriate management may be obtained through volumetric analysis that is not readily available, even in retrospect, from standard two-dimensional analysis of the same raw data. Unlike computed tomographic angiography (CTA), this analysis does not require intravenous contrast administration, as generally it is not the geometry of the contrast column that is of interest to the surgeon. In addition to potential direct cost savings, this has important implications for a population of patients who frequently have impaired renal function, and who may require other contrast examinations (e.g. coronary angiography) as part of the same preinterventional workup. We have also demonstrated the ability of volumetric analysis to derive more accurate and reliable quantitative measurements of structures with complex three-dimensional morphology.
Standard two-dimensional analysis of transverse CT images are vulnerable to errors of reliability, bias, and accuracy. The relatively high reliability of standard two- dimensional measurements from transverse CT images may be due to the fact that alignment of the patient along the longitudinal axis in serial scans is highly repeatable. Also, there is substantial agreement between readings and amongst readers upon the image representing the greatest apparent aneurysmal diameter. An approach similar to that employed by Lederle et al.28 in the large VA cooperative study tends to increase this reliability by regulating the algorithm applied to obtain a measurement. Difficulty arises in the absence of a "gold standard," since any inherent bias which affects all readers approximately equally will be obscured. In other words, the standard system of measurement may be highly internally reliable but significantly inaccurate.
Such was the case in our study of phantom aneurysms. Readers agreed fairly closely upon measured diameters, but comparison with the known true diameter of the phantoms demonstrated a substantial bias which affected both readers similarly, causing them to "agree" upon the wrong value. We believe that a important component of this bias is represented by the combined effects of vessel tortuosity and cross-sectional eccentricity. Alone, the tortuosity of a true cylinder would be accurately accounted for by measuring the shortest diameter of the oval cross- section in the transverse image. However, as we have discovered in a large proportion of the patients so far evaluated by volumetric analysis [unpublished data], eccentricity is far more common than previously appreciated. Additionally, the severity of vessel tortuosity and its effect upon measurements also tends to be underestimated or missed entirely on standard transverse CT images. If the cumulative inaccuracy generated by these influences is no larger than the errors measured under the idealized conditions of our phantom study, errors in estimated diameter may range up to a centimeter or more. Since repair of an aneurysm is usually undertaken for a perceived interval increase in diameter of greater than 5 mm in a six month period, the magnitude of such unrecognized errors may materially affect clinical decisions. Standards for normal aortic diameters, as well as for outcomes analysis of stable, growing, and ruptured aneurysms, are also based upon this invalid system of measurement. Inaccurate stratification of patients at risk for morbidity and mortality from abdominal aortic aneurysms, resulting from an invalid system of measurement, may increase both human and economic costs related to this disease process.
We believe that the increases in accuracy and precision, and amelioration of bias, demonstrated by volumetric analysis in the evaluation of aneurysm phantoms should translate well to the clinical arena. This system of measurement deserves further study and validation in human subjects. If the quality of analysis is sustained, longitudinal outcomes research based upon this system should help better define patients at risk from AAA, as well as defining appropriate strategies for management more sensitive to cost without eliminating benefit.
We deeply appreciate our colleagues' generous contributions of support, time, and expertise, especially Kevin Berbaum, PhD, Jim Olson, Walter Seaman, PhD, William Higgins, PhD, Natalie Oppel, Joe Reinhardt, Ron Stahlberg, Neil D'Souza, and De Frei. This project was funded in part by a contract from the National Library of Medicine: N01-LM-4-3511 US PHS.
Papers |
DPI Homepage |
VIDA |
NLM |
Contact Us |
Search