If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Discrepancy between maximum diameters obtained with two-dimensional ultrasound and computed tomography (CT) after endovascular aneurysm repair (EVAR) is well known. The maximal diameter is ideally measured perpendicular to the centerline, a methodology so far only feasible with three-dimensional (3D) CT and magnetic resonance angiography (MRA). We aimed to investigate the agreement between 3D ultrasound and 3D CT and to determine reproducibility measures.
Prospective study comparing 3D ultrasound with 3D CT in 124 consecutive patients seen 3 or 12 month after EVAR.
Replacing 2D with 3D ultrasound, the mean difference was improved from 6.0 mm to −1.3 mm (p < .001), and the range of variability was reduced from 9.4 mm to 6.6 mm (p = .009) using 3D CT as the gold standard. The mean difference between 3D ultrasound and 3D CT maximum diameter of the residual sac was −1.3 mm with upper and lower limits of agreement of 5.2 mm and −7.9 mm, respectively. Reproducibility measures of 3D ultrasound were ±4 mm.
3D ultrasound correlate significantly better to 3D CT than the currently used 2D ultrasound method when assessing maximum diameter of the residual sac after EVAR, and reproducibility measures were within clinical acceptable values.
Ultrasonic assessment of abdominal aortic aneurysms has traditionally been confined to two-dimensional (2D) imaging. Determination of a centerline and diameter assessment perpendicular to this line has been the gold standard by 3D CT. We present a novel 3D ultrasound method capable of measuring the maximum diameter of the residual sac perpendicular to the centerline herby reducing the well known discrepancy.
The size of the residual sac is an important parameter after endovascular aortic repair (EVAR) since expansion could imply incomplete aneurysm exclusion and treatment failure, whereas shrinkage and stability indicate operative success.
Besides being the gold standard in assessment of residual sac diameter, dynamic computed tomographic angiography (CTA) has a high sensitivity for endoleak detection and is therefore the most frequently used modality in EVAR surveillance.
EVAR surveillance protocols based on ultrasound are nevertheless emerging because of the risk of contrast-induced nephropathy and radiation associated with the use of CTA and the acceptable ability of ultrasound to detect endoleaks.
2D CT measurements, however, can be made from the entire circumferential wall of the aneurysms, whereas the ultrasound reflection is strongest from the anterior and posterior aortic wall. Care must be taken with the expansion of the residual sac, especially if surveillance is changed from a CT-based to an ultrasound-based protocol. 3D ultrasound technology has recently emerged as an alternative, and likewise CTA technology. The principle of 3D reconstruction allows the maximum diameter perpendicular to the centerline of the aneurysm to be measured.
We aimed to determine the accuracy of 3D ultrasound and 2D ultrasound using 3D CTA as the gold standard, and, secondly, to determine the reproducibility of 3D ultrasound.
Materials and Methods
Patients and study design
All consecutive patients scheduled for standard EVAR surveillance at 3- or 12-month follow-up in the period from August 2011 to July 2012 were prospectively enrolled into the study after giving informed consent. The study was approved by the Regional Ethical Committee (H-2–2011–016).
All patients were initially treated at our institution using the same EVAR device (Zenith stent-grafts, Cook Medical Inc., Bloomington, IN, USA).
All patients had CTA, plain abdominal X-ray, and an ultrasound investigation including 2D examination and 3D acquisitions. For each patient four maximal aneurysm diameters were assessed and compared: 2D ultrasound, 3D ultrasound, 2D CT and 3D CT (Fig. 1). We used the maximal diameter obtained by 3D CT as the gold standard.
If patients were seen at both 3- and 12-month follow-up during the study period we only included examinations obtained at the first visit.
A subset of the patients included in this study has previously been described in a methodology study introducing volume estimation of the residual sac.
Patients were not instructed to fast but had 10 minutes of rest before the examinations. One physician (KB) experienced in vascular ultrasound (>1,000 vascular ultrasound investigations) performed all investigations with the patient in supine position using the ultrasound system (Philips iU22 Ultrasound System, Philips Healthcare, Bothell, WA, USA).
First, the 3D ultrasound acquisition was performed followed by the traditional 2D ultrasound examination including endoleak check. Then the same operator performed a second 3D ultrasound acquisition used for intraoperator assessment. Finally, in cases used for interoperator assessment, a senior consultant (JE) experienced in vascular ultrasound (>15 years' experience in vascular ultrasound) performed a comparative 3D ultrasound acquisition.
The 3D ultrasound acquisition was performed with a commercially available 3D transducer (X6-1 xMATRIX, Philips Healthcare). The maximum cross-sectional diameter was determined and assisted by dual-plane imaging keeping the boundaries of the residual sac within the scan field of view. 3D acquisition consists of an electronic sweep acquiring multiple images simultaneously in both longitudinal and transverse directions. During breath-hold, the acquisition was completed within approximately 1 second, with the transducer in a stable firm position. No diameter information potentially biasing the 2D measurement was displayed during the 3D ultrasound acquisition.
The 2D ultrasound investigations were all performed blinded to the results of 2D CT, 3D CT, and 3D ultrasound diameter measurement with a 5-MHz curved array transducer (C5-1, Philips Healthcare). As part of a standard ultrasound EVAR surveillance scan, the abdominal aorta was interrogated from the superior mesenteric artery to the level of the aortic bifurcation. Identifying the greatest section of the residual, the transducer was tilted from side to side so that the maximal diameter was measured from the leading edge of the adventitia of the anterior wall to the leading edge of the adventitia on the far wall in transverse view (Fig. 1A).
In order to correct for obliquity in the sagittal plane, the measurement obtained in the transverse plane had to correlate within 3 mm of the measurement performed in longitudinal imaging.
Biphasic acquisition (unenhanced and contrast-enhanced with bolus tracking) was performed using a helical 64-slice CT scanner (Toshiba Medical Systems Ltd, Crawley, UK). Detector configuration was 0.5 × 64 (collimation = 32 mm) with a pitch of 0.8. A bolus dose of 80 mL of non-ionic iodinated contrast medium (Iohexol 350 mg I/mL, Omnipaque; GE Healthcare Denmark A/S, Copenhagen, Denmark) was injected into an antecubital vein at a rate of 3 mL/second. Scan reconstructions were performed with a slice thickness and increment of 3 mm.
The 2D CT maximal diameter was measured from outer to outer circumferential wall in any direction by an experienced radiologist (MT) (>1,000 EVAR CT evaluations), blinded to the results of ultrasound and 3D CT, using a PACS system (Agfa Impax 5.2, Agfa-Gevaert NV, Mortsel, Belgium) allowing assessment in both the native axial CT slices and in the multiplanar reconstructions (sagittal and coronal plane) (Fig. 1B). In cases of tortuosity the diameter was measured perpendicular to the direction of tortuosity.
3D reconstruction and diameter assessment for ultrasound and CT
Paired ultrasound acquisitions and contrast-enhanced CT images were analysed in dedicated 3D interactive software (AAA_prototype, version 1.0, Medisys, Philips Research, Suresnes, France) with an interpolated interval of at least 14 days.
On the ultrasound and CT acquisitions, the residual sac, including the stent-graft, was semi-automatically delineated using the inner vessel wall with the 3D interactive segmentation technique.
It took between 15 and 30 minutes to segment the 3D ultrasound acquisition.
For both modalities, the centerline of the residual sac was automatically generated between two manually located extremities in the most proximal and distal locations of the sac (Fig. 1E). Finally, the maximum 3D ultrasound and 3D CT diameter perpendicular to the centerline was generated from inner to inner vessel wall.
Inter- and intraoperator variability
3D ultrasound interoperator variability
Patients (n = 22) were selected by their concurrent presence on weekdays when both operators (KB and JE), mutually blinded, were available and were rescanned a few minutes after by the second operator (JE), who afterwards independently analysed his own 3D acquisitions in the 3D software.
3D ultrasound intraoperator variability
In order to make the intraoperator assessment as mutually blinded as possible, the conventional EVAR surveillance protocol, including 2D ultrasound scan and physical examination, was interpolated resulting in around 15–25 minutes between the first and second 3D ultrasound acquisition.
The two 3D acquisitions were then analysed in the 3D interactive software with time intervals of at least 14 days.
3D CT interoperator variability
The interoperator range of variability (ROV) of the maximal CT diameter perpendicular to the centerline was determined between two operators (AL and KB) familiar with the 3D software, who independently assessed the CT scans of all patients (n = 28) enrolled during the first 3 months of the study.
The maximal residual sac diameter was measured with four imaging techniques and was expressed as the mean ± 95% confidence interval (95% CI) given by 1.96 × standard deviation of the mean (SD). To compare means we used the paired Student's t test.
Comparison of residual sac measurements were represented in a Bland–Altman plot, where discrepancy of paired measurement on the same subject were plotted against the average outcome, showing the mean difference and the upper and lower limits of agreements given by the mean difference ± the ROV. The ROV was defined as 1.96 × SD of the mean difference.
When evaluating a method's accuracy, it is not sufficient only to assess the mean difference from the gold standard, which may be different simply due to different physical properties associated with CTA and ultrasound. Hence, observed differences in ROV between methods were tested for homogeneity using Levene's test.
All statistical analyses were performed by SAS v. 9.3 (SAS Institute Inc., Cary, NC, USA).
Patients and technical success rate
In total, 124 patients had paired ultrasound and CT examinations at 3 months (n = 72) or at 12 months (n = 52) following EVAR. Insonation was impossible due to bowel gas in four patients, and in one patient having an aortic–bi-iliac aneurysm, the abdominal component could not be isolated from the iliac aneurysms making the 3D ultrasound reconstruction impossible, leaving an overall technical success rate of 3D ultrasound of 96% (119/124). Patients having aortic–iliac aneurysms (n = 5) and aortic–bi-iliac aneurysms (n = 3) were included if it was possible to isolate the abdominal component in the 3D reconstructions. We observed no patients having claustrophobia or being unable to cooperate, no patients refused to participate in this investigation, and no adverse events were observed after CTA or ultrasound.
We enrolled 13 females and 106 males with a mean age of 78 years (range 57–90 years) and median body mass index of 24 (range 16–41) kg/cm2. The mean maximum preoperative aneurysm diameter obtained with 2D CT was 63.8 (95% CI 44.9–82.6). There were 12 patients with aneurysms < 55 mm, of which four had symptomatic non-ruptured aneurysms; six female patients women and two male patients had asymptomatic aneurysms of 54 mm and 53 mm with more than 5 mm of growth since last visit. The EVAR device was bifurcated in 87% (n = 104), aorto-uni-iliac in 12% (n = 14), and a fenestrated endograft was used in 1% (n = 1) of the cases. One patient seen at 12 months after Palmaz-stent deployment due to proximal sealing defect had a growing sac. Growth and type I leak were equivalently identified on CTA and ultrasound. Type II leaks were present in 14% (n = 17) of the patients, of which 15 had a stable sac and two had sac shrinkage. No patients were observed with the combination of type II leak and increasing maximum diameter.
Maximal residual sac diameter
2D and 3D ultrasound and 2D CT compared with 3D CT
The mean maximum residual sac diameter obtained with 3D CT, 2D CT, 3D ultrasound, and 2D ultrasound was 60.8 (95% CI 38.9–82.7) mm, 61.5 (95% CI 39.9–83.1), 62.2 (95% CI 39.1–85.2), and 54.8 (95% CI 34.6–75.1), respectively.
The mean differences ± ROV of 2D CT, 3D ultrasound, and 2D ultrasound compared with 3D CT were −0.7 ± 6.0 mm, –1.3 ± 6.6 mm, and 6.0 ± 9.4 mm, respectively. Only the mean maximum diameter obtained with 2D ultrasound was more variable than the mean diameter obtained with 3D CT (p = .002), being the gold standard (Table 1 and Fig. 2). Furthermore, when measuring the maximum diameter using 3D ultrasound instead of 2D ultrasound compared with 3D CT, the ROV was reduced to 6.6 mm from 9.4 mm (p = 0.009).
Table 1The mean residual sac diameter obtained by the four imaging methods and the associated mean difference from three-dimensional (3D) computed tomography (CT), the gold standard.
Mean diameter (mm)
Mean difference ±range of variability from 3D CT (mm)
When comparing 3D ultrasound and 2D CT using 3D CT as the gold standard, neither the mean difference nor ROV of −1.3 ± 6.6 mm performed with 3D ultrasound was different from that performed with 2D CT of −0.7 ± 6.0 mm (mean difference, p = .13) (ROV, p = .053).
When analysing 3D CT versus 3D ultrasound and increasing mean aneurysm diameter, a minimal negative slope of the linear regression line was observed (slope = −0.0535, p = .0487). In contrast, the linear regression line of 3D CT versus 2D ultrasound had a positive slope (slope = 0.0823, p = .0490) (Fig. 2).
Linear regression showed that the absolute difference between paired 3D ultrasound and 3D CT measurements slightly deteriorated with increasing preoperative diameter (slope = 0.052, p = .0146) and the diameter (slope = 0.054, p = .0176) obtained at time of inclusion. The comparability deteriorated by 0.5 mm with an increase of 1 cm of either the preoperative diameter or the diameter obtained at inclusion. Thus, having preoperative residual sacs of 5.5 cm and 7.5 cm the absolute difference between paired 3D ultrasound and 3D CT measures increased from 2.4 mm to 3.4 mm.
Comparing the maximum diameters obtained with 3D ultrasound and 3D CT the discrepancy was less than 5 mm in 87% (n = 103) of the cases. The remaining 13% (n = 16) of patients were further assessed. The main reasons for the larger discrepancy, was imprecise segmentation of the lateral vessel wall leading to overestimation of the ultrasound measure (n = 8), shadowing of the posterolateral vessel wall due to a close positioned stent-graft (n = 1), poor image quality in the distal and proximal range of the scan field leading to erroneously defined endolimb extremities (n = 4), retroperitoneal fibrosis leading to overestimation of the ultrasound assessment (n = 1), inclusion of the cava vein (n = 1), and compression of the residual sac against the corpus vertebrae by the transducer (n = 1).
2D and 3D ultrasound compared with 2D CT
When 2D ultrasound was compared with 2D CT, representing the most commonly used imaging methods in clinical practice, we observed a mean difference of 6.7 ± 9.8 mm between the two. Replacing 2D ultrasound with 3D ultrasound the mean difference was reduced to −0.6 mm (p < .001) but the ROV of 9.2 mm was unchanged (p = .687).
Inter- and intraoperator reproducibility
3D ultrasound reproducibility
The mean maximal diameter difference between two operators performing and assessing their own acquisition was −0.1 mm, which was not different from the mean difference obtained between two acquisitions performed and assessed by the same operator of 0.7 mm (p = .312) (Fig. 3).
The 3D ultrasound interoperator ROV was 4.4 mm, which was not different from the 3D ultrasound intraoperator ROV of 3.9 mm (p = .239).
Linear regression line of differences showed that neither the interoperator (slope = −0.1121, p = .0664.) nor the intraoperator (slope = 0.0696, p = .2572) variability was influenced by increasing residual sac size.
3D CT reproducibility
The mean difference between the first and second operator evaluating the same CT scan (n = 28) was −0.1 mm and the ROV was 2.1 mm, which was smaller than the interoperator ROV of 3D ultrasound (p < .001). Linear regression showed that the 3D CT interoperator variability was not influenced by increasing residual sac size (p = .6947).
The maximal diameter of the residual sac after EVAR can be measured more accurately using 3D ultrasound instead of 2D ultrasound according to our study. Not only was the ROV reduced by one-third but the mean difference from the gold standard, 3D CT, was close to zero. Finally, the reproducibility of 3D ultrasound was approximately 4 mm, which we believe is acceptable for clinical use.
To our knowledge, no studies assessing the maximum abdominal aortic aneurysm diameter using 3D ultrasound has been published. Until now, studies evaluating the agreement between ultrasound and CT have been confined to the use of 2D ultrasound. These studies reported that the mean difference between CTA and ultrasound ranged from 1.3 mm to 3.9 mm and the variability ranged from 10 mm to 12.2 mm.
The reported range of reproducibility measures of 2D ultrasound was, however, restricted to anteroposterior measurement. Since the main part of previous conducted reproducibility studies regarding 2D ultrasound were motivated by the set up of screening programmes, mainly small aneurysms were included in these studies where incorrect angle of insonation is less critical than measurement of large aneurysms.
In this context, it is important to notice several challenging factors only present in EVAR patients. First, the residual sacs are often large; however, linear regression did not show that increasing residual sac size had a significant effect on the 3D ultrasound reproducibility measures. On the other hand, the difference between 3D ultrasound and the gold standard was slightly increasing with increasing residual sac size. Second, aorta and the residual sac are remodelled after endovascular therapy, presenting an irregular and thrombotic post-interventional sac and a centrally located EVAR device making insonation of the posterior vessel wall challenging.
Ultrasound investigations of patients undergoing EVAR surveillance programmes have until recently been confined to 2D assessment, while centerline-determined diameter only has been possible using CTA and a dedicated 3D workstation, being the gold standard.
The demonstrated better agreement between 3D ultrasound and 2D and 3D CT when using 3D technology has the potential to reduce the risk of neglecting real expansion of the residual sac in ultrasound-based EVAR surveillance.
Recent developments in radiation-reduced dual-energy CTAs have been reported and MR protocols are shown to be more sensitive than CTA.
Despite this, institutions with trained ultrasound personal intend to replace CTA with ultrasound in their EVAR surveillance protocol due to the risk of contrast-induced nephropathy associated with CTA. Moreover, using ultrasound the physician has the possibility to inform the patient immediately in the outpatient's clinic (one-visit clinic), and. dependent on local arrangements, the availability and associated cost may favour ultrasound instead of MR/CTA. Moreover, some EVAR devices with stainless steel, as used in this study, are MR incompatible.
For clinical use, however, 3D ultrasound still has some challenges to overcome. The segmentation process involves a learning curve and is time consuming explaining why 3D ultrasound at our institution is still considered a research tool. Nevertheless, future improvement of the software is expected, including automatic and faster 3D segmentation, and within some years we foresee 3D ultrasound to be an important complementary imaging modality for accurate size estimation in “radiation-free EVAR surveillance”. This, however, demands solid longitudinal follow-up data and our future results may confirm or reject that 3D ultrasound does not miss any growing residual sacs that were otherwise diagnosed by CTA.
Since most institutions use CTA, especially within the first year, the essential question is whether ultrasound can replace CTA in EVAR surveillance which includes endoleak detection for which we consider the image quality of the matrix transducer to be inadequate. Contrast-enhanced 2D ultrasound is promising, and therefore, in parallel to this study, we are trying to show that endoleaks can be detected equivalently compared with CTA. Till then, we still consider conventional 2D ultrasound in trained hands to be an effective image modality checking uncomplicated EVAR cases without growth or endoleaks seen after 1 year. Abdominal aortic aneurysm is a potential life-threatening surgical condition where size is a critical determinant, not only regarding EVAR surveillance. In the surveillance of small asymptomatic aneurysms, it is well known that 2D ultrasound tends to underestimate aneurysm size, which is why aneurysms approaching 5 cm in diameter on ultrasound often are referred to CTA.
Accurate correlation between these image modalities would rationalize the need for CTA. Whether 3D ultrasound imaging provides better reproducibility measures for small, native abdominal aortic aneurysms is a matter of future research, but we would expect that the absence of the stent-graft would improve the signal-to-noise ratio and reflection of the vessel wall and thus improve the results.
This prospective single-centre study benefits by a homogeneous population in terms of anatomical classification, conformed logistics, that is the same type of EVAR device inserted, ultrasound investigations performed by one operator, and CTA scans assessed by a single operator. Poor 2D images, disturbed by bowel gas or calcification, may to some extent be compensated by the high number of frames and different beam angles that the new 3D ultrasound technique provides. Obviously, we recognize that the same high technical success rate may be difficult to reproduce in a clinical setting.
The 3D CT reproducibility was superior to the 3D ultrasound reproducibility. This is presumably in part because the same CTA acquisition was used twice in contradiction to the 3D ultrasound reproducibility examinations where two different acquisitions were used. To make two CTA acquisitions is simply not ethical due to the risk of renal impairment and the radiation dose exposure to the patient. We, nevertheless, recognize that a better definition of the lateral wall on CTA than ultrasound is of great importance, emphasizing that delineation of the lateral vessel was by far the most critical step involved using 3D ultrasound. This is supported by the outliers, where we observed imprecise lateral wall segmentation and thereby lack of a smooth 3D shape.
On the other hand, the integration of the lateral wall in 3D ultrasound is one of the strengths and may very well explain the improved agreement with 2D and 3D CT. In the transverse plane, the reflection from the lateral vessel wall in 2D ultrasound can be poor. For these reasons 2D ultrasound measures are, opposed to 2D CT measures, more valid when obtained from the anterior to the posterior wall. In cases where the maximum diameter was located in the coronal plane from “side to side”, the operator tried to adjust for this by tilting the transducer (see Fig. 1A) but the ideal coronal plane could be difficult to achieve using 2D ultrasound. This detail may lead to less optimal agreement with 3D CT in cases where the maximum diameter is in the coronal plane. The poorer reflection from the lateral vessel is no different using 3D ultrasound, but the integration of multiple ultrasound frames instead of one and reconstruction of a centerline tends to adjust for this, which means that the cross-section evaluated on 3D CT and ultrasound are more alike.
Since the operators were vascular surgeons and angiologists familiar with the purpose of the study, it may have been relevant to assess the performance of an independent radiologist, which we did not.
Finally, the 3D ultrasound and 3D CT segmentation depends on the inner vessel wall whereas the radiologist use outer-to-outer wall for traditional 2D CT measurements. In 2D ultrasound, assessment was made from the leading edge of the adventitia on the anterior wall to the leading edge of the adventitia on the far wall. The reason for using inner vessel when comparing 3D CT and 3D ultrasound was that the 3D software depends on distinct differences in greyscale; differences being most evident over the boundary “aortic-lumen/inner-wall”. Using inner to inner wall in 2D ultrasound would result in an even smaller mean maximal diameter, and therefore and even worse agreement with 3D CT could be expected.
In conclusion, we introduce a new 3D ultrasound method to measure the maximum diameter of the residual sac after EVAR perpendicular to the centerline with improved agreement with 3D CT and acceptable reproducibility.
The Danish Heart Foundation funded this project.
Conflict of Interest
Co-author Sillesen H has received a research grant and honorarium from Philips Ultrasound .
Management of abdominal aortic aneurysms clinical practice guidelines of the European society for vascular surg.
To submit a comment for a journal article, please use the space above and note the following:
We will review submitted comments as soon as possible, striving for within two business days.
This forum is intended for constructive dialogue. Comments that are commercial or promotional in nature, pertain to specific medical cases, are not relevant to the article for which they have been submitted, or are otherwise inappropriate will not be posted.
We require that commenters identify themselves with names and affiliations.