CLINICAL STUDY
Aortic Arch Vessel Geometries and Deformations in Patients with Thoracic Aortic Aneurysms and Dissections Ga-Young Suh, PhD, Ramin E. Beygui, MD, Dominik Fleischmann, MD, and Christopher P. Cheng, PhD
ABSTRACT Purpose: To quantify aortic arch geometry and in vivo cardiac-induced and respiratory-induced arch translations and arch branch angulations using three-dimensional geometric modeling techniques. Materials and Methods: Scanning with electrocardiogram-gated computed tomography angiography during inspiratory and expiratory breath holds was performed in 15 patients (age, 64 y ⫾ 14) with thoracic aortic aneurysms or dissections. From the lumen models, centerlines of the thoracic aorta, brachiocephalic artery, left common carotid artery, and left subclavian artery and their branching ostia positions were quantified. Three-dimensional translation of vessel ostia, branching angles, and their changes secondary to cardiac pulsation and respiration were computed. Results: During expiration, all ostia translated rightward from systole to diastole (P o .035). Regardless of cardiac phase, all ostia translated posteriorly and superiorly from inspiration to expiration (P o .05). Respiration induced greater posterior and superior translations than cardiac pulsation (P o .03). The left common carotid artery branch angled significantly more toward the aortic arch compared with the brachiocephalic artery and left subclavian artery (P o .03). No significant changes in branching angle were found from systole to diastole or inspiration to expiration. Conclusions: In patients with thoracic aortic aneurysms or dissections, the thoracic aortic arch translated significantly secondary to inspiration and expiration and to a lesser extent secondary to cardiac pulsation. Insignificant branching angle changes suggest that the aortic arch and its branch origins move predominantly in unison.
ABBREVIATIONS BA = brachiocephalic artery, LCCA = left common carotid artery, LSA = left subclavian artery, 3D = three-dimensional
Aneurysms and dissections of the descending thoracic aorta are commonly treated by endovascular approaches (1,2), whereas diseases of the ascending aorta and the aortic arch are primarily addressed by open surgical repair with prosthetic grafts (3). Ongoing developments in endovascular techniques and stent graft technology have led to more attention focused on less invasive
From the Departments of Surgery (G.-Y.S., C.P.C.), Cardiothoracic Surgery (R.E.B.), and Radiology (D.F.), Stanford University, 300 Pasteur Drive, Suite H3600, Stanford, CA 94305-5642. Received February 3, 2014; final revision received May 28, 2014; accepted June 9, 2014. Address correspondence to G.-Y.S.; E-mail: [email protected] From the SIR 2014 Annual Meeting. None of the authors have identified a conflict of interest. & SIR, 2014 J Vasc Interv Radiol 2014; 25:1903–1911 http://dx.doi.org/10.1016/j.jvir.2014.06.012
alternatives, and the application of endovascular repair has expanded tremendously for the descending aorta (4– 7). However, similar treatment for the aortic arch is complicated by short landing zones and complex geometries of the brachiocephalic artery (BA), left common carotid artery (LCCA), and left subclavian artery (LSA) (8). To preserve the patency for these arch vessels while repairing the diseased aortic arch, endovascular devices require branched components for these arch vessels. Although multiple studies have employed branched endovascular grafts to treat the ascending aorta and arch, no devices designed for this procedure have been approved by the U.S. Food and Drug Administration (9,10). The incidence of device failure and endoleaks after treatment indicates the need for endograft design improvement (11,12). Design challenges of endograft for thoracic aortic aneurysm or dissection repair include high blood pressure and pulsatile hemodynamic force
1904
’
Aortic Arch Geometries and Deformations
owing to proximity to the heart, vulnerable and inhomogeneous aortic anatomy with challenging landing zones, and aortic motion caused by respiration (13). To design and test branched endografts with the promise of long-term performance, precise understanding of arch vessel geometries and deformations is required (13,14). We propose quantification techniques to measure arch vessel geometry and deformations using computed tomography (CT) angiography–based three-dimensional (3D) geometric modeling techniques and quantify cardiac-induced and respiratory-induced aortic arch translations and branching angle changes of the BA, LCCA, and LSA in patients with thoracic aortic aneurysms and dissections.
MATERIALS AND METHODS Patient Recruitment We prospectively enrolled 15 nonconsecutive patients (mean age, 64 y; range, 40–88 y; 9 men and 6 women) in this study. To screen the candidates, the following inclusion criteria were used: (a) existence of chronic aortic dissections or aortic aneurysms of the thoracic aorta and (b) native descending aorta without surgical or endovascular repair history. The following exclusion criteria were used: (a) diagnosis of Marfan or EhlersDanlos syndrome, (b) prior history of aortic valve repair with David procedure or surgical root fixation, (c) prior history of arch branch stent placement, and (d) poor breathing capability or severe pulmonary disease. The first exclusion criterion was introduced to exclude patients with genetic disorders of connective tissue, which may possess different mechanical properties. The second and third criteria were to exclude the nonnative loading and mechanical properties at the aortic root and arch vessels. The fourth criteria was added because the candidates were required to hold their breath during image acquisition. The study protocol was approved by the institutional review board, and written informed consent was received from each patient.
Image Acquisition CT angiography images of the chest were acquired using 64-channel CT technology with retrospective electrocardiogram gating employing a dual-source CT scanner (SOMATOM Definition; Siemens Healthcare, Erlangen, Germany) or a single-source CT scanner (LightSpeed VCT; GE Healthcare, Waukesha, Wisconsin). The first acquisition (phase 1) was acquired during an inspiration breath hold using 100–190 mL (based on the patient’s body weight) of nonionic contrast medium (ISOVUE 300 or ISOVUE 370; Bracco Diagnostics, Inc, Princeton, New Jersey) injected at a rate of 4–5 mL/s. The scan delay time was determined by automated bolus triggering (SmartPrep; GE Healthcare) (15,16). Subsequently, the patients were instructed to breathe normally for 30
Suh et al
’
JVIR
seconds, and then a second scan was performed during an expiration breath hold without additional contrast injection (phase 2). Images for phases 1 and 2 were acquired with 120-kVp and 100-kVp tube potentials, respectively, with 0.6-mm to 0.625-mm collimation and 0.2–0.33 pitch factor. Resultant volume CT index was approximately 38 mGy and 24 mGy for phases 1 and 2, respectively. Axial CT images were reconstructed with a section thickness of 1 mm for the dual-source CT scan and 1.25 mm for the single-source CT scan, and a reconstruction field of view between 30 30 cm and 40 40 cm. CT scans were acquired with retrospective electrocardiogram gating allowing for the reconstruction of time-resolved CT data sets at 10% intervals of the cardiac cycle. End-diastolic images were reconstructed at 70% of the R-R interval, and end-systolic images were reconstructed at 40% of the R-R interval. Four physiologic modes were used for further analysis: inspiration þ diastole, inspiration þ systole, expiration þ diastole, and expiration þ systole.
Modeling and Centerline Extraction Using the custom software SimVascular (Open Source Medical Software Corporation, San Diego, California), CT images were employed to construct 3D lumen models and centerline paths of the thoracic aorta, BA, LCCA, and LSA (Fig 1a–e) (17,18). First, an initial vessel path was created by hand-picking lumen centers and connecting these points with interpolated cubic splines. Along the initial path, orthogonal lumen cross sections were segmented with an automated twodimensional level set segmentation technique (18) with a spacing of about half of the vessel radius. The set of segmentations was lofted to create a 3D lumen model. Finally, true vessel centerline paths were defined by computing the mathematical centroids for each lumen contour, connecting the centroids through the vessel, and smoothing the centroid-based path with a 0.1-mm point interpolation and an optimized number of Fourier modes (19). The output of this modeling was the 3D coordinates of the centroid-based centerline paths of the thoracic aorta, BA, LCCA, and LSA. Using MATLAB (MathWorks, Natick, Massachusetts), the centerline path coordinates were loaded to compute arch branch positions and branching angles for each vessel at all physiologic modes as well as their differences.
Quantification of Aortic Arch Translation The branch positions of the BA, LCCA, and LSA on the aortic arch were tracked for each physiologic mode. Cardiac-induced aortic arch translation was computed by subtracting the position at systole from the position at diastole. Respiratory-induced aortic arch translation was computed by subtracting the position at inspiration from the position at expiration.
Volume 25
’
Number 12
’
December
’
2014
1905
Figure 1. Lumen segmentation and centerline path extraction. (a) CT angiography images were loaded into the custom software SimVascular. (b) Vessel paths were constructed by connecting the hand-picked lumen centers along the thoracic aorta, BA, LCCA, LSA, and coronary arteries. (c) Along the vessel path, two-dimensional level set segmentations were performed to segment the lumen boundaries of the vessels. A 3D model was formed by lofting the contours to a two-dimensional shell surface. (d) Mathematical centroids were found from the segmentation contours. (e) Centerline paths were constructed by connecting the multiple centroids with cubic splines and reconstructed by optimized Fourier smoothing.
Quantification of Branching Angle The branching angle was defined as the angle between the aortic vector along the centerline path of the thoracic aorta and branch vessel vectors along the centerline paths of the BA, LCCA, and LSA. The aortic vector was defined by two points on the aortic centerline—the first at the level of the branch vessel ostium and the second 10 mm distal on the centerline. The branch vessel vectors were defined from the branch ostium to three different locations: 10 mm, 20 mm, and 30 mm distal along the branch vessel centerline (Fig 2). A higher branching angle indicates that the vessel is angled further away from the aortic arch (more orthogonal), whereas a lower branching angle indicates that the vessel is angled closer to the aortic arch (more parallel). A cardiac-induced change in the branching angle was computed by subtracting the angle at systole from the angle at diastole. A respiratory-induced change in the branching angle was computed by subtracting the angle at inspiration from the angle at expiration.
Statistical Analysis Population data were reported as mean ⫾ SD. Twotailed paired t tests were used for comparison between vessels and between physiologic modes. For multiple comparisons, the significance threshold (P ¼ .05) was adjusted by Bonferroni-Holm correction (20). All calculations were performed with the statistical package in Excel (Microsoft, Redmond, Washington).
RESULTS Patient Recruitment The thoracic aorta and arch vessels are depicted with 3D-rendered CT angiography images and lumen models for 15 patients (Fig 3). A detailed description of aortic pathology for these patients is provided in Table 1.
Figure 2. Calculation of branching angle of BA, LCCA, and LSA. Branching angle was defined as the angle between two vectors, one along the centerline path of the thoracic aorta and one along the path of each of BA, LCCA, and LSA. The thoracic aorta vector was defined by the point 10 mm distal to the branch ostium on the thoracic aorta centerline. The branch vessels vectors were defined by points 10 mm, 20 mm, and 30 mm distal to the ostia along the branch vessel centerlines.
Patients 1, 2, 3, and 4 had ascending aortic aneurysms without surgical repair of the aorta. Patients 5, 6, 7, 9, 12, 13, 14, and 15 had chronic aortic dissections at the ascending aorta repaired with prosthetic grafts, and some had residual dissection at the aortic arch and descending aorta. Patients 10 and 11 had chronic aortic dissections at the ascending aorta and arch repaired with elephant trunk procedures (21). For patients with residual dissections in the thoracic aorta and arch vessels, the lumen models included both whole lumen, bounded by the most external wall boundaries, and true lumen, bounded by the internal wall boundaries. All models presented in Figure 3 were constructed from the contrast
1906
’
Aortic Arch Geometries and Deformations
Suh et al
’
JVIR
Figure 3. 3D rendering of thoracic aorta and arch vessels from CT angiography (left image) and lumen model built with custom software SimVascular (right image) for patients with ascending aortic aneurysms or aortic dissections at end diastole during inspiration breath hold. Four patients had ascending aortic aneurysms without surgical repair of the aorta (patients 1–4). Eight patients had chronic aortic dissections at the ascending aorta repaired with prosthetic grafts (patients 5–7, 9, 12–15). Two patients had chronic aortic dissections at the ascending aorta and arch repaired with elephant trunk procedures (patients 10, 11). True lumen with residual dissection was modeled for visualization purposes (dark gray). Whole lumen was modeled and used for geometric quantification (light gray). (Available in color online at www.jvir.org.)
CT images at inspiration þ diastole. Although not shown in Figure 3, models for inspiration þ systole, expiration þ diastole, and expiration þ systole were also constructed. Quantification of geometries and deformations was performed with the whole-lumen aortic arch and arch vessels.
Visual Comparison between Physiologic Modes For visual comparison between inspiration þ diastole, inspiration þ systole, expiration þ diastole, and expiration þ systole, lumen models of patient 2 were coregistered and are presented in Figure 4a–e. For the visual comparison, lumen models in four physiologic modes were loaded to the single visual window in the custom software. These models were constructed from the CT angiography scans with consistent field of view and coordinate system. From systole to diastole during expiration breath-hold, the aortic
root translated slightly superiorly, and the aortic arch translated rightward (Fig 4b). From inspiration to expiration at both diastole and systole, the thoracic aorta and arch vessels shifted posteriorly and superiorly (Fig 4c, d). Greater translation occurred secondary to respiration compared with cardiac contraction (Fig 4e).
Aortic Arch Translation Translation of arch vessel ostia on the aortic arch in leftto-right, posterior-to-anterior, and inferior-to-superior directions and 3D translations are shown in Table 2. From systole to diastole, the aortic arch translated rightward significantly with expiration (P o .035) at the BA, LCCA, and LSA ostia. From inspiration to expiration, the aortic arch translated posteriorly and superiorly at all three ostial locations (P o .05). Respiration (inspiration þ diastole to expiration þ diastole) induced greater translation than cardiac pulsation (inspiration þ diastole to inspiration þ systole) in posterior (P o .03)
Volume 25
’
Number 12
’
December
’
2014
1907
Table 1 . Disease Profile of 15 Patients with Thoracic Aortic Aneurysms or Dissections Patient No.
Ascending Aorta
Aortic Arch
Descending Aorta
BA
LCCA
LSA
1
A, D
—
—
—
—
—
2
A
—
—
—
—
—
3 4
A A
A —
A —
— —
— —
— —
5
RD
D
A, D
D
D
D
6 7
RD RD
D D
D A, D
— D
D —
D —
8
—
D
D
—
D
—
9 10
RD RD
D RD
D A, D
— —
— —
D —
11
RD
RD
A, D
—
—
D
12 13
RD RD
D —
D —
— —
— —
— —
14
RD
—
A
—
—
—
15
RD
—
—
D
—
—
A ¼ aortic aneurysm; A, D ¼ dissected aneurysm; BA ¼ brachiocephalic artery; D ¼ visible residual dissection; LCCA ¼ left common carotid artery; LSA ¼ left subclavian artery; RD ¼ surgically repaired dissection.
and superior (P o .03) directions at all three ostia. In 3D translations, respiratory-induced translation was consistently greater than cardiac-induced translation (P o .024).
Branching Angles and Changes with Cardiac Pulsation and Respiration Branching angles between the aortic arch and branch vessels were averaged for the 15 patients (Table 3). The branching angle of the LCCA was significantly less than the branching angle of the BA and LSA in all modes for the 30-mm branch vector and some modes for the 10mm and 20-mm branch vectors (P o .03). No significant differences were found between the BA and LSA. Changes in the branching angle from systole to diastole and from inspiration to expiration were averaged for 15 patients (Table 4). In each vessel, no statistically significant changes were found from systole to diastole or inspiration to expiration. Branching angle changes were not significantly different between the three arch vessels or between branch vector lengths.
23). In addition, lumen centerlines were formed by connecting these computed centroids and undergoing optimal Fourier smoothing, which has been shown to have an accuracy of 0.2 mm (19).
Aortic Arch Translation Cardiac pulsation and respiration resulted in significant translation of the aortic arch and branch vessels but with distinctively different direction and magnitude. From systole to diastole, the aortic arch shifted rightward during expiration (P o .035). The rightward translation is likely due to the release of the aortic root away from the heart apex during diastolic relaxation (24). From inspiration to expiration, significant posterior and superior translations were observed at all arch ostia (P o .05), likely owing to abdominal contraction and diaphragmatic shift (25). In most cases, respiration induced greater translation than cardiac pulsation in the posterior and superior directions (P o .03).
Branching Angles of the Arch Vessels DISCUSSION Quantification Methods The present study uses quantitative modeling techniques to measure aortic arch translation and branching angles of the arch vessels. The fidelity of these measurements depends on the accuracy and precision of fiducial marker selection (eg, branch ostia). The fiducial markers in this study depend on accurate centerline extraction, so centerline computation is the most critical step in this process. We used vessel lumen segmentation and computation of centroids of these lumens to provide a much more accurate and user-independent method of determining lumen centers compared with manual selection (19,22,
The BA, LCCA, and LSA exhibited different branching geometries. In most cases, the branching angle of the LCCA was significantly lower than the branching angles of other vessels, owing to angulation toward the arch, which can be seen in most of the patients (Fig 3). The angulation of the LCCA makes sense because it branches from the aortic arch and crosses over to the left side of the body. Conversely, the BA does not exhibit this angled geometry because the BA branches from the right side of the body and extends superiorly before branching into the right subclavian and carotid arteries. The differences in arch branching angles need to be considered when designing or choosing the conformability of branched endografts.
1908
’
Aortic Arch Geometries and Deformations
Suh et al
’
JVIR
Figure 4. One example of vessel motion secondary to cardiac pulsation and respiration by coregistering lumen models to compare (a) diastole (gray) versus systole (red) during inspiration, (b) diastole (blue) versus systole (yellow) during expiration, (c) inspiration (grey) versus expiration (blue) at end diastole, (d) inspiration (red) versus expiration (yellow) at end systole, and (e) all four physiologic modes. A, R, and S indicate anterior, rightward, and superior axes for each model view. Note the translation of the thoracic aorta and arch vessels in the posterior and superior directions with expiration in (c) and (d). All models, constructed from the consistent field-of-view CT angiography scans, were loaded to the single visual window of the custom software so that 3D translations of one model to another were visualized. (Available in color online at www.jvir.org.)
Branching Angle Changes Despite significant arch translation, changes in branching angles of the arch vessels were not significant. Although nonsignificant branching angle changes may result from aortic arch translation, it seems that the aortic arch and branch vessels move predominantly in unison in the patients of this study. This finding suggests that aortic arch endografts with branched components may not be subjected to aggressive cyclic bending and angulation at the arch branches.
Comparison with Previous Studies The aortic translation data from this study, owing to cardiac and respiratory influences, corroborate well with
other studies (22,25,26). We report average cardiacinduced 3D arch translations of 0.3–7.8 mm, whereas Weber et al (22) reported 3D translations of 0.4–4.0 mm. One reason our results reflect a greater range may be because we tracked translations of the BA, LCCA, and LSA ostia, whereas Weber et al tracked only the aortic lumen centroid just distal to the LSA. We report average respiratory-induced 3D arch translations of 1.5– 17.9 mm, whereas Maxim et al (26) reported translations of 1.2–14.7 mm. We tracked three points on the aortic arch, whereas Weber et al (22) tracked only a single point just distal to the LSA. Our findings of greater thoracic aortic translation with respiration compared with cardiac pulsation are consistent with a previous study (25).
Volume 25
’
Number 12
’
December
’
2014
1909
Table 2 . Cardiac-Induced and Respiratory-Induced 3D Aortic Arch Translation (mm) at BA, LCCA, and LSA Branch Points BA (mm)
LCCA (mm)
LSA (mm)
0.2 ⫾ 1.8
0.2 ⫾ 1.3
Left-to-right translation Diastole-systole
Inspiration
0.3 ⫾ 1.8
Expiration-inspiration
Expiration Diastole
0.7 ⫾ 1.2* 1.3 ⫾ 2.7
Systole
0.9 ⫾ 2.8
0.4 ⫾ 2.8
0.5 ⫾ 2.2
Inspiration
0.4 ⫾ 1.5
0.0 ⫾ 1.4
0.1 ⫾ 1.6
Expiration
0.8 ⫾ 1.2*
1.1 ⫾ 1.4*
0.6 ⫾ 1.7
Diastole Systole
3.9 ⫾ 4.6* 3.4 ⫾ 4.3*
3.3 ⫾ 4.0* 2.3 ⫾ 3.9*
2.8 ⫾ 3.7* 2.2 ⫾ 3.7*
Posterior-to-anterior translation Diastole-systole Expiration-inspiration
0.8 ⫾ 1.0* 0.6 ⫾ 2.9
1.0 ⫾ 0.9* 0.2 ⫾ 2.6
Inferior-to-superior translation Diastole-systole
Inspiration Expiration
0.7 ⫾ 1.7 0.6 ⫾ 2.3
0.3 ⫾ 1.7 1.0 ⫾ 2.5
0.8 ⫾ 1.3* 0.2 ⫾ 2.3
Expiration-inspiration
Diastole
2.6 ⫾ 3.9*
3.2 ⫾ 3.6*
3.4 ⫾ 3.7*
Systole Inspiration
2.8 ⫾ 3.8* 2.4 ⫾ 1.7
2.5 ⫾ 3.3* 2.2 ⫾ 1.7
4.0 ⫾ 3.4* 2.0 ⫾ 1.5
Diastole-systole
Expiration
2.5 ⫾ 1.7
2.9 ⫾ 1.8
2.8 ⫾ 1.5
Expiration-inspiration
Diastole Systole
6.7 ⫾ 4.4 6.6 ⫾ 3.9
6.3 ⫾ 4.2 5.6 ⫾ 3.5
6.1 ⫾ 3.9 6.4 ⫾ 3.1
3D translation
Note.–Values are mean ⫾ SD. The significance threshold (P o .05) was adjusted by Bonferroni-Holm correction for multiple comparisons. BA ¼ brachiocephalic artery, LCCA ¼ left common carotid artery, LSA ¼ left subclavian artery, 3D ¼ three-dimensional *Statistically significant translation from one physiologic mode to another. Table 3 . Branching Angle (degrees) of BA, LCCA, and LSA at Four Physiologic Modes with Branch Vessel Vectors Defined by Points 10, 20, and 30 mm Distal to the Branch Ostia
Branch vector 10 mm distal Inspiration
BA (mm)
LCCA (mm)
LSA (mm)
Diastole
58.9 ⫾ 18.7
52.8 ⫾ 15.4*
65.6 ⫾ 19.5*
Systole
58.6 ⫾ 20.6
51.8 ⫾ 15.2*
65.5 ⫾ 17.4*
Diastole Systole
60.4 ⫾ 21.2 60.0 ⫾ 18.8
55.1 ⫾ 17.5 54.7 ⫾ 14.3*
66.0 ⫾ 20.1 65.8 ⫾ 19.6*
Inspiration
Diastole Systole
56.9 ⫾ 20.5† 55.0 ⫾ 22.5
48.2 ⫾ 16.3*† 46.6 ⫾ 15.6*
63.1 ⫾ 19.5* 63.1 ⫾ 22.5*
Expiration
Diastole
59.6 ⫾ 26.9
50.3 ⫾ 18.8*
63.8 ⫾ 23.0*
Systole
57.5 ⫾ 23.0
50.3 ⫾ 14.9
63.1 ⫾ 21.2
Diastole
59.9 ⫾ 25.7†
44.7 ⫾ 18.3*†
61.9 ⫾ 23.4*
Systole Diastole
†
58.0 ⫾ 26.7 63.2 ⫾ 31.6†
43.7 ⫾ 17.0*† 47.3 ⫾ 21.0*†
61.5 ⫾ 20.9* 62.8 ⫾ 23.8*
Systole
60.9 ⫾ 27.9†
47.4 ⫾ 18.5*†
62.2 ⫾ 22.6*
Expiration Branch vector 20 mm distal
Branch vector 30 mm distal Inspiration Expiration
Note.–Values are mean ⫾ SD. The significance threshold (P o .05) was adjusted by Bonferroni-Holm correction for multiple comparisons. No significant differences were found between BA and LSA. No significant differences were found between the physiologic modes or between branch vector lengths. BA ¼ brachiocephalic artery, LCCA ¼ left common carotid artery, LSA ¼ left subclavian artery. *Significant difference between LCCA and LSA. † Significant difference between BA and LCCA.
Limitations This study has limitations that could be addressed in future work. First, the small sample size and heterogeneous cohort limit the power of statistical analysis. In
addition, the repaired ascending aorta in some of the patients may exaggerate aortic arch motion because of the shorter distance from the heart to the arch and the greater transmission of cardiac motion to the arch because of the
1910
’
Aortic Arch Geometries and Deformations
Suh et al
’
JVIR
Table 4 . Differences in Branching Angle (degrees) of BA, LCCA, and LSA between Four Physiologic Modes with Branch Vessel Vectors Defined by Points 10, 20, and 30 mm Distal to the Branch Ostia BA (mm)
LCCA (mm)
LSA (mm)
Inspiration
0.3 ⫾ 4.2
1.0 ⫾ 4.3
0.1 ⫾ 4.8
Expiration
0.5 ⫾ 6.7
0.3 ⫾ 7.8
0.2 ⫾ 7.0
Diastole Systole
1.6 ⫾ 7.9 1.4 ⫾ 5.0
2.3 ⫾ 4.7 3.0 ⫾ 5.6
0.5 ⫾ 6.3 0.3 ⫾ 7.1
Diastole-systole
Inspiration Expiration
1.9 ⫾ 3.6 2.1 ⫾ 7.7
1.6 ⫾ 3.0 0.0 ⫾ 7.9
0.0 ⫾ 4.1 0.7 ⫾ 5.7
Expiration-inspiration
Diastole
2.7 ⫾ 9.0
2.1 ⫾ 4.6
0.7 ⫾ 5.9
Systole
2.5 ⫾ 5.1
3.7 ⫾ 6.0
0.0 ⫾ 6.8
Diastole-systole
Inspiration
1.9 ⫾ 3.3
1.0 ⫾ 2.6
0.4 ⫾ 3.9
Expiration-inspiration
Expiration Diastole
2.4 ⫾ 7.2 3.3 ⫾ 8.4
0.1 ⫾ 7.1 2.6 ⫾ 5.4
0.6 ⫾ 5.5 0.9 ⫾ 5.8
Systole
2.9 ⫾ 5.4
3.7 ⫾ 6.6
0.7 ⫾ 7.3
Branch vector 10 mm distal Diastole-systole Expiration-inspiration Branch vector 20 mm distal
Branch vector 30 mm distal
Note.–Values are mean ⫾ SD. The significance threshold (P o .05) was adjusted by Bonferroni-Holm correction for multiple comparisons. No significant differences were found between the physiologic modes or between vessels. BA ¼ brachiocephalic artery, LCCA ¼ left common carotid artery, LSA ¼ left subclavian artery.
stiffness of the surgical graft. Second, motion artifacts secondary to fast systolic motion, irregular heart rate, and muscle movement during breath hold cause image blurring that affects the precision of lumen segmentation. Third, although we analyzed only snapshots at end systole, diastole, inspiration, and expiration, a full analysis of cardiac-resolved and respiratory-resolved images could be done to understand the progression during a complete physiologic event (27). Finally, cardiac-induced and respiratory-induced deformations of the thoracic aorta were not quantified separately but could dramatically affect the dynamics of the aortic arch (28,29). In conclusion, this study quantified the 3D translations of the aortic arch, branching angles of the arch vessels, and their changes secondary to cardiac pulsation and respiration. The aortic arch translated significantly rightward from systole to diastole and posteriorly and superiorly from inspiration to expiration. Respiratoryinduced translation was significantly greater than cardiac-induced translation. Although branching angles of the BA, LCCA, and LSA were different from each other, none of them deformed significantly with cardiac or respiratory influences. The geometric and deformation data of the arch branch vessels may aid in treatment planning, implantation technique, and design and evaluation of branched thoracic aortic endografts. In addition, these quantitative methods can be used in other anatomies and to investigate changes of vessel geometry and compliance before and after endovascular repair.
ACKNOWLEDGMENT This work was supported by a research gift from Medtronic Inc. We thank Lior Molvin, Daisha Marsh, Christoph
Zorich, Monglan Duong, and Fatin Alkhadra for help with CT imaging and Riley Marangi for help with modeling. We also thank all the patients for their participation.
REFERENCES 1. Desai ND, Burtch K, Moser W, et al. Long-term comparison of thoracic endovascular aortic repair (TEVAR) to open surgery for the treatment of thoracic aortic aneurysms. J Thorac Cardiovasc Surg 2012; 144: 604–611. 2. Preventza O, Cervera R, Cooley DA, et al. Acute type I aortic dissection: traditional versus hybrid repair with antegrade stent delivery to the descending thoracic aorta. J Thorac Cardiovasc Surg 2014; 148: 119–125. 3. Tian DH, Wan B, Di Eusanio M, Black D, Yan TD. A systematic review and meta-analysis on the safety and efficacy of the frozen elephant trunk technique in aortic arch surgery. Ann Cardiothorac Surg 2013; 2:581–591. 4. Bonser RS, Ranasinghe AM, Loubani M, et al. Evidence, lack of evidence, controversy, and debate in the provision and performance of the surgery of acute type A aortic dissection. J Am Coll Cardiol 2011; 58: 2455–2474. 5. Lopera J, Patino JH, Urbina C, et al. Endovascular treatment of complicated type-B aortic dissection with stent-grafts: midterm results. J Vasc Interv Radiol 2003; 14:195–203. 6. Lee M, Lee DY, Kim MD, et al. Outcomes of endovascular management for complicated chronic type B aortic dissection: effect of the extent of stent graft coverage and anatomic properties of aortic dissection. J Vasc Interv Radiol 2013; 24:1451–1460. 7. Moon MC, Morales JP, Greenberg RK. The aortic arch and ascending aorta: are they within the endovascular realm? Semin Vasc Surg 2007; 20:97–107. 8. Chuter TAM, Hiramoto JS, Chang C, et al. Branched stent-grafts: will these become the new standard? J Vasc Interv Radiol 2008; 19:S57–S62. 9. Greenberg RK, O’Neill S, Walker E, et al. Endovascular repair of thoracic aortic lesions with the Zenith TX1 and TX2 thoracic grafts: intermediateterm results. J Vasc Surg 2005; 41:589–596. 10. Nordon IM, Hinchliffe RJ, Morgan R, Loftus IM, Jahangiri M, Thomson MM. Progress in endovascular management of type A dissection. Eur J Vasc Endovasc Surg 2012; 44:406–410. 11. Bavaria JE, Appoo JJ, Makaroun MS, Verter J, Yu Z-F, Mitchell RS. Endovascular stent grafting versus open surgical repair of descending thoracic aortic aneurysms in low-risk patients: a multicenter comparative trial. J Thorac Cardiovasc Surg 2007; 133:369–377.
Volume 25
’
Number 12
’
December
’
2014
12. Cheng D, Martin J, Shennib H, et al. Endovascular aortic repair versus open surgical repair for descending thoracic aortic disease: a systematic review and meta-analysis of comparative studies. J Am Coll Cardiol 2010; 55:986–1001. 13. Weidman JM, Desai M, Iftekhar A, et al. Engineering goals for future thoracic endograft—how can we make them more effective? Prog Cardiovasc Disease 2013; 56:92–102. 14. Zarins CK, Taylor CA. Endovascular device design in the future: transformation from trial and error to computational design. J Endovasc Ther 2009; 16:I12–I21. 15. Silverman PM, Roberts S, Tefft MC, et al. Helical CT of the liver: clinical application of an automated computer technique, SmartPrep, for obtaining images with optimal contrast enhancement. AJR Am J Roentgenol 1995; 165:73–78. 16. Fleischmann D. How to design injection protocols for multiple detector-row CT angiography (MDCTA). Eur Radiol Suppl 2005; 15: E60–E65. 17. Wang KC. Level Set Methods for Computational Prototyping with Application to Hemodynamic Modeling. Ph.D. Thesis. Stanford, CA: Stanford University Press; 2001. 18. Wilson N, Wang K, Dutton RW, Taylor CA. A software framework for creating patient specific geometric models from medical imaging data for simulation based medical planning of vascular surgery. Lect Notes Comput Sci 2001; 2208:449–456. 19. Choi G, Cheng CP, Wilson NM, Taylor CA. Methods for quantifying three-dimensional deformation of arteries due to pulsatile and nonpulsatile forces: implications for the design of stents and stent grafts. Ann Biomed Eng 2009; 37:14–33. 20. Holm A. A simple sequentially rejective multiple test procedure. Scand J Stat 1979; 6:65–70.
1911
21. Schepens MA. Aortic arch replacement: the conventional ‘elephant trunk’ technique. Multimed Man Cardiothorac Surg 2007; 2007(102): mmcts.2006.002006. 22. Weber TF, Ganten MK, Böckler D, Geisbüsch P, Kauzor HU, von TenggKobligk H. Heartbeat-related displacement of the thoracic aorta in patients with chronic aortic dissection type B: quantification by dynamic CTA. Eur J Radiol 2009; 72:483–488. 23. Goel VR, Greenberg RK, Greenberg DP. Mathematical analysis of DICOM CT datasets: can endograft sizing be automated for complex anatomy? J Vasc Surg 2008; 47:1306–1312. 24. Scott AD, Keegan J, Firmin DN. Motion in cardiovascular MR imaging. Radiology 2009; 250:331–351. 25. Weber TF, Tetzlaff R, Rengier F, et al. Respiratory displacement of the thoracic aorta: physiological phenomenon with potential implications for thoracic endovascular repair. Cardiovasc Intervent Radiol 2009; 32: 658–665. 26. Maxim PG, Loo BW Jr, Shirazi H, Thorndyke B, Luxton G, Le QT. Quantification of motion of different thoracic locations using fourdimensional computed tomography: implications for radiotherapy planning. Int J Radiat Oncol Biol Phys 2007; 69:1395–1401. 27. Weber TF, Muller T, Biesdorf A, et al. True four-dimensional analysis of thoracic aortic displacement and distension using model-based segmentation of computed tomography angiography. Int J Cardiovasc Imaging 2014; 30:185–194. 28. Morrison TM, Choi G, Zarins CK, Taylor CA. Circumferential and longitudinal cyclic strain of the human thoracic aorta: age-related changes. J Vasc Surg 2009; 49:1029–1036. 29. Carrascosa P, Capunay C, Deviggiano A, et al. Thoracic aorta cardiaccycle related dynamic changes assessed with a 256-slice CT scanner. Cardiovasc Diagn Ther 2013; 3:125–128.
Aortic Arch Vessel Geometries and Deformations in Patients with Thoracic Aortic Aneurysms and Dissections Ga-Young Suh, PhD, Ramin E. Beygui, MD, Dominik Fleischmann, MD, and Christopher P. Cheng, PhD
ABSTRACT Purpose: To quantify aortic arch geometry and in vivo cardiac-induced and respiratory-induced arch translations and arch branch angulations using three-dimensional geometric modeling techniques. Materials and Methods: Scanning with electrocardiogram-gated computed tomography angiography during inspiratory and expiratory breath holds was performed in 15 patients (age, 64 y ⫾ 14) with thoracic aortic aneurysms or dissections. From the lumen models, centerlines of the thoracic aorta, brachiocephalic artery, left common carotid artery, and left subclavian artery and their branching ostia positions were quantified. Three-dimensional translation of vessel ostia, branching angles, and their changes secondary to cardiac pulsation and respiration were computed. Results: During expiration, all ostia translated rightward from systole to diastole (P o .035). Regardless of cardiac phase, all ostia translated posteriorly and superiorly from inspiration to expiration (P o .05). Respiration induced greater posterior and superior translations than cardiac pulsation (P o .03). The left common carotid artery branch angled significantly more toward the aortic arch compared with the brachiocephalic artery and left subclavian artery (P o .03). No significant changes in branching angle were found from systole to diastole or inspiration to expiration. Conclusions: In patients with thoracic aortic aneurysms or dissections, the thoracic aortic arch translated significantly secondary to inspiration and expiration and to a lesser extent secondary to cardiac pulsation. Insignificant branching angle changes suggest that the aortic arch and its branch origins move predominantly in unison.
ABBREVIATIONS BA = brachiocephalic artery, LCCA = left common carotid artery, LSA = left subclavian artery, 3D = three-dimensional
Aneurysms and dissections of the descending thoracic aorta are commonly treated by endovascular approaches (1,2), whereas diseases of the ascending aorta and the aortic arch are primarily addressed by open surgical repair with prosthetic grafts (3). Ongoing developments in endovascular techniques and stent graft technology have led to more attention focused on less invasive
From the Departments of Surgery (G.-Y.S., C.P.C.), Cardiothoracic Surgery (R.E.B.), and Radiology (D.F.), Stanford University, 300 Pasteur Drive, Suite H3600, Stanford, CA 94305-5642. Received February 3, 2014; final revision received May 28, 2014; accepted June 9, 2014. Address correspondence to G.-Y.S.; E-mail: [email protected] From the SIR 2014 Annual Meeting. None of the authors have identified a conflict of interest. & SIR, 2014 J Vasc Interv Radiol 2014; 25:1903–1911 http://dx.doi.org/10.1016/j.jvir.2014.06.012
alternatives, and the application of endovascular repair has expanded tremendously for the descending aorta (4– 7). However, similar treatment for the aortic arch is complicated by short landing zones and complex geometries of the brachiocephalic artery (BA), left common carotid artery (LCCA), and left subclavian artery (LSA) (8). To preserve the patency for these arch vessels while repairing the diseased aortic arch, endovascular devices require branched components for these arch vessels. Although multiple studies have employed branched endovascular grafts to treat the ascending aorta and arch, no devices designed for this procedure have been approved by the U.S. Food and Drug Administration (9,10). The incidence of device failure and endoleaks after treatment indicates the need for endograft design improvement (11,12). Design challenges of endograft for thoracic aortic aneurysm or dissection repair include high blood pressure and pulsatile hemodynamic force
1904
’
Aortic Arch Geometries and Deformations
owing to proximity to the heart, vulnerable and inhomogeneous aortic anatomy with challenging landing zones, and aortic motion caused by respiration (13). To design and test branched endografts with the promise of long-term performance, precise understanding of arch vessel geometries and deformations is required (13,14). We propose quantification techniques to measure arch vessel geometry and deformations using computed tomography (CT) angiography–based three-dimensional (3D) geometric modeling techniques and quantify cardiac-induced and respiratory-induced aortic arch translations and branching angle changes of the BA, LCCA, and LSA in patients with thoracic aortic aneurysms and dissections.
MATERIALS AND METHODS Patient Recruitment We prospectively enrolled 15 nonconsecutive patients (mean age, 64 y; range, 40–88 y; 9 men and 6 women) in this study. To screen the candidates, the following inclusion criteria were used: (a) existence of chronic aortic dissections or aortic aneurysms of the thoracic aorta and (b) native descending aorta without surgical or endovascular repair history. The following exclusion criteria were used: (a) diagnosis of Marfan or EhlersDanlos syndrome, (b) prior history of aortic valve repair with David procedure or surgical root fixation, (c) prior history of arch branch stent placement, and (d) poor breathing capability or severe pulmonary disease. The first exclusion criterion was introduced to exclude patients with genetic disorders of connective tissue, which may possess different mechanical properties. The second and third criteria were to exclude the nonnative loading and mechanical properties at the aortic root and arch vessels. The fourth criteria was added because the candidates were required to hold their breath during image acquisition. The study protocol was approved by the institutional review board, and written informed consent was received from each patient.
Image Acquisition CT angiography images of the chest were acquired using 64-channel CT technology with retrospective electrocardiogram gating employing a dual-source CT scanner (SOMATOM Definition; Siemens Healthcare, Erlangen, Germany) or a single-source CT scanner (LightSpeed VCT; GE Healthcare, Waukesha, Wisconsin). The first acquisition (phase 1) was acquired during an inspiration breath hold using 100–190 mL (based on the patient’s body weight) of nonionic contrast medium (ISOVUE 300 or ISOVUE 370; Bracco Diagnostics, Inc, Princeton, New Jersey) injected at a rate of 4–5 mL/s. The scan delay time was determined by automated bolus triggering (SmartPrep; GE Healthcare) (15,16). Subsequently, the patients were instructed to breathe normally for 30
Suh et al
’
JVIR
seconds, and then a second scan was performed during an expiration breath hold without additional contrast injection (phase 2). Images for phases 1 and 2 were acquired with 120-kVp and 100-kVp tube potentials, respectively, with 0.6-mm to 0.625-mm collimation and 0.2–0.33 pitch factor. Resultant volume CT index was approximately 38 mGy and 24 mGy for phases 1 and 2, respectively. Axial CT images were reconstructed with a section thickness of 1 mm for the dual-source CT scan and 1.25 mm for the single-source CT scan, and a reconstruction field of view between 30 30 cm and 40 40 cm. CT scans were acquired with retrospective electrocardiogram gating allowing for the reconstruction of time-resolved CT data sets at 10% intervals of the cardiac cycle. End-diastolic images were reconstructed at 70% of the R-R interval, and end-systolic images were reconstructed at 40% of the R-R interval. Four physiologic modes were used for further analysis: inspiration þ diastole, inspiration þ systole, expiration þ diastole, and expiration þ systole.
Modeling and Centerline Extraction Using the custom software SimVascular (Open Source Medical Software Corporation, San Diego, California), CT images were employed to construct 3D lumen models and centerline paths of the thoracic aorta, BA, LCCA, and LSA (Fig 1a–e) (17,18). First, an initial vessel path was created by hand-picking lumen centers and connecting these points with interpolated cubic splines. Along the initial path, orthogonal lumen cross sections were segmented with an automated twodimensional level set segmentation technique (18) with a spacing of about half of the vessel radius. The set of segmentations was lofted to create a 3D lumen model. Finally, true vessel centerline paths were defined by computing the mathematical centroids for each lumen contour, connecting the centroids through the vessel, and smoothing the centroid-based path with a 0.1-mm point interpolation and an optimized number of Fourier modes (19). The output of this modeling was the 3D coordinates of the centroid-based centerline paths of the thoracic aorta, BA, LCCA, and LSA. Using MATLAB (MathWorks, Natick, Massachusetts), the centerline path coordinates were loaded to compute arch branch positions and branching angles for each vessel at all physiologic modes as well as their differences.
Quantification of Aortic Arch Translation The branch positions of the BA, LCCA, and LSA on the aortic arch were tracked for each physiologic mode. Cardiac-induced aortic arch translation was computed by subtracting the position at systole from the position at diastole. Respiratory-induced aortic arch translation was computed by subtracting the position at inspiration from the position at expiration.
Volume 25
’
Number 12
’
December
’
2014
1905
Figure 1. Lumen segmentation and centerline path extraction. (a) CT angiography images were loaded into the custom software SimVascular. (b) Vessel paths were constructed by connecting the hand-picked lumen centers along the thoracic aorta, BA, LCCA, LSA, and coronary arteries. (c) Along the vessel path, two-dimensional level set segmentations were performed to segment the lumen boundaries of the vessels. A 3D model was formed by lofting the contours to a two-dimensional shell surface. (d) Mathematical centroids were found from the segmentation contours. (e) Centerline paths were constructed by connecting the multiple centroids with cubic splines and reconstructed by optimized Fourier smoothing.
Quantification of Branching Angle The branching angle was defined as the angle between the aortic vector along the centerline path of the thoracic aorta and branch vessel vectors along the centerline paths of the BA, LCCA, and LSA. The aortic vector was defined by two points on the aortic centerline—the first at the level of the branch vessel ostium and the second 10 mm distal on the centerline. The branch vessel vectors were defined from the branch ostium to three different locations: 10 mm, 20 mm, and 30 mm distal along the branch vessel centerline (Fig 2). A higher branching angle indicates that the vessel is angled further away from the aortic arch (more orthogonal), whereas a lower branching angle indicates that the vessel is angled closer to the aortic arch (more parallel). A cardiac-induced change in the branching angle was computed by subtracting the angle at systole from the angle at diastole. A respiratory-induced change in the branching angle was computed by subtracting the angle at inspiration from the angle at expiration.
Statistical Analysis Population data were reported as mean ⫾ SD. Twotailed paired t tests were used for comparison between vessels and between physiologic modes. For multiple comparisons, the significance threshold (P ¼ .05) was adjusted by Bonferroni-Holm correction (20). All calculations were performed with the statistical package in Excel (Microsoft, Redmond, Washington).
RESULTS Patient Recruitment The thoracic aorta and arch vessels are depicted with 3D-rendered CT angiography images and lumen models for 15 patients (Fig 3). A detailed description of aortic pathology for these patients is provided in Table 1.
Figure 2. Calculation of branching angle of BA, LCCA, and LSA. Branching angle was defined as the angle between two vectors, one along the centerline path of the thoracic aorta and one along the path of each of BA, LCCA, and LSA. The thoracic aorta vector was defined by the point 10 mm distal to the branch ostium on the thoracic aorta centerline. The branch vessels vectors were defined by points 10 mm, 20 mm, and 30 mm distal to the ostia along the branch vessel centerlines.
Patients 1, 2, 3, and 4 had ascending aortic aneurysms without surgical repair of the aorta. Patients 5, 6, 7, 9, 12, 13, 14, and 15 had chronic aortic dissections at the ascending aorta repaired with prosthetic grafts, and some had residual dissection at the aortic arch and descending aorta. Patients 10 and 11 had chronic aortic dissections at the ascending aorta and arch repaired with elephant trunk procedures (21). For patients with residual dissections in the thoracic aorta and arch vessels, the lumen models included both whole lumen, bounded by the most external wall boundaries, and true lumen, bounded by the internal wall boundaries. All models presented in Figure 3 were constructed from the contrast
1906
’
Aortic Arch Geometries and Deformations
Suh et al
’
JVIR
Figure 3. 3D rendering of thoracic aorta and arch vessels from CT angiography (left image) and lumen model built with custom software SimVascular (right image) for patients with ascending aortic aneurysms or aortic dissections at end diastole during inspiration breath hold. Four patients had ascending aortic aneurysms without surgical repair of the aorta (patients 1–4). Eight patients had chronic aortic dissections at the ascending aorta repaired with prosthetic grafts (patients 5–7, 9, 12–15). Two patients had chronic aortic dissections at the ascending aorta and arch repaired with elephant trunk procedures (patients 10, 11). True lumen with residual dissection was modeled for visualization purposes (dark gray). Whole lumen was modeled and used for geometric quantification (light gray). (Available in color online at www.jvir.org.)
CT images at inspiration þ diastole. Although not shown in Figure 3, models for inspiration þ systole, expiration þ diastole, and expiration þ systole were also constructed. Quantification of geometries and deformations was performed with the whole-lumen aortic arch and arch vessels.
Visual Comparison between Physiologic Modes For visual comparison between inspiration þ diastole, inspiration þ systole, expiration þ diastole, and expiration þ systole, lumen models of patient 2 were coregistered and are presented in Figure 4a–e. For the visual comparison, lumen models in four physiologic modes were loaded to the single visual window in the custom software. These models were constructed from the CT angiography scans with consistent field of view and coordinate system. From systole to diastole during expiration breath-hold, the aortic
root translated slightly superiorly, and the aortic arch translated rightward (Fig 4b). From inspiration to expiration at both diastole and systole, the thoracic aorta and arch vessels shifted posteriorly and superiorly (Fig 4c, d). Greater translation occurred secondary to respiration compared with cardiac contraction (Fig 4e).
Aortic Arch Translation Translation of arch vessel ostia on the aortic arch in leftto-right, posterior-to-anterior, and inferior-to-superior directions and 3D translations are shown in Table 2. From systole to diastole, the aortic arch translated rightward significantly with expiration (P o .035) at the BA, LCCA, and LSA ostia. From inspiration to expiration, the aortic arch translated posteriorly and superiorly at all three ostial locations (P o .05). Respiration (inspiration þ diastole to expiration þ diastole) induced greater translation than cardiac pulsation (inspiration þ diastole to inspiration þ systole) in posterior (P o .03)
Volume 25
’
Number 12
’
December
’
2014
1907
Table 1 . Disease Profile of 15 Patients with Thoracic Aortic Aneurysms or Dissections Patient No.
Ascending Aorta
Aortic Arch
Descending Aorta
BA
LCCA
LSA
1
A, D
—
—
—
—
—
2
A
—
—
—
—
—
3 4
A A
A —
A —
— —
— —
— —
5
RD
D
A, D
D
D
D
6 7
RD RD
D D
D A, D
— D
D —
D —
8
—
D
D
—
D
—
9 10
RD RD
D RD
D A, D
— —
— —
D —
11
RD
RD
A, D
—
—
D
12 13
RD RD
D —
D —
— —
— —
— —
14
RD
—
A
—
—
—
15
RD
—
—
D
—
—
A ¼ aortic aneurysm; A, D ¼ dissected aneurysm; BA ¼ brachiocephalic artery; D ¼ visible residual dissection; LCCA ¼ left common carotid artery; LSA ¼ left subclavian artery; RD ¼ surgically repaired dissection.
and superior (P o .03) directions at all three ostia. In 3D translations, respiratory-induced translation was consistently greater than cardiac-induced translation (P o .024).
Branching Angles and Changes with Cardiac Pulsation and Respiration Branching angles between the aortic arch and branch vessels were averaged for the 15 patients (Table 3). The branching angle of the LCCA was significantly less than the branching angle of the BA and LSA in all modes for the 30-mm branch vector and some modes for the 10mm and 20-mm branch vectors (P o .03). No significant differences were found between the BA and LSA. Changes in the branching angle from systole to diastole and from inspiration to expiration were averaged for 15 patients (Table 4). In each vessel, no statistically significant changes were found from systole to diastole or inspiration to expiration. Branching angle changes were not significantly different between the three arch vessels or between branch vector lengths.
23). In addition, lumen centerlines were formed by connecting these computed centroids and undergoing optimal Fourier smoothing, which has been shown to have an accuracy of 0.2 mm (19).
Aortic Arch Translation Cardiac pulsation and respiration resulted in significant translation of the aortic arch and branch vessels but with distinctively different direction and magnitude. From systole to diastole, the aortic arch shifted rightward during expiration (P o .035). The rightward translation is likely due to the release of the aortic root away from the heart apex during diastolic relaxation (24). From inspiration to expiration, significant posterior and superior translations were observed at all arch ostia (P o .05), likely owing to abdominal contraction and diaphragmatic shift (25). In most cases, respiration induced greater translation than cardiac pulsation in the posterior and superior directions (P o .03).
Branching Angles of the Arch Vessels DISCUSSION Quantification Methods The present study uses quantitative modeling techniques to measure aortic arch translation and branching angles of the arch vessels. The fidelity of these measurements depends on the accuracy and precision of fiducial marker selection (eg, branch ostia). The fiducial markers in this study depend on accurate centerline extraction, so centerline computation is the most critical step in this process. We used vessel lumen segmentation and computation of centroids of these lumens to provide a much more accurate and user-independent method of determining lumen centers compared with manual selection (19,22,
The BA, LCCA, and LSA exhibited different branching geometries. In most cases, the branching angle of the LCCA was significantly lower than the branching angles of other vessels, owing to angulation toward the arch, which can be seen in most of the patients (Fig 3). The angulation of the LCCA makes sense because it branches from the aortic arch and crosses over to the left side of the body. Conversely, the BA does not exhibit this angled geometry because the BA branches from the right side of the body and extends superiorly before branching into the right subclavian and carotid arteries. The differences in arch branching angles need to be considered when designing or choosing the conformability of branched endografts.
1908
’
Aortic Arch Geometries and Deformations
Suh et al
’
JVIR
Figure 4. One example of vessel motion secondary to cardiac pulsation and respiration by coregistering lumen models to compare (a) diastole (gray) versus systole (red) during inspiration, (b) diastole (blue) versus systole (yellow) during expiration, (c) inspiration (grey) versus expiration (blue) at end diastole, (d) inspiration (red) versus expiration (yellow) at end systole, and (e) all four physiologic modes. A, R, and S indicate anterior, rightward, and superior axes for each model view. Note the translation of the thoracic aorta and arch vessels in the posterior and superior directions with expiration in (c) and (d). All models, constructed from the consistent field-of-view CT angiography scans, were loaded to the single visual window of the custom software so that 3D translations of one model to another were visualized. (Available in color online at www.jvir.org.)
Branching Angle Changes Despite significant arch translation, changes in branching angles of the arch vessels were not significant. Although nonsignificant branching angle changes may result from aortic arch translation, it seems that the aortic arch and branch vessels move predominantly in unison in the patients of this study. This finding suggests that aortic arch endografts with branched components may not be subjected to aggressive cyclic bending and angulation at the arch branches.
Comparison with Previous Studies The aortic translation data from this study, owing to cardiac and respiratory influences, corroborate well with
other studies (22,25,26). We report average cardiacinduced 3D arch translations of 0.3–7.8 mm, whereas Weber et al (22) reported 3D translations of 0.4–4.0 mm. One reason our results reflect a greater range may be because we tracked translations of the BA, LCCA, and LSA ostia, whereas Weber et al tracked only the aortic lumen centroid just distal to the LSA. We report average respiratory-induced 3D arch translations of 1.5– 17.9 mm, whereas Maxim et al (26) reported translations of 1.2–14.7 mm. We tracked three points on the aortic arch, whereas Weber et al (22) tracked only a single point just distal to the LSA. Our findings of greater thoracic aortic translation with respiration compared with cardiac pulsation are consistent with a previous study (25).
Volume 25
’
Number 12
’
December
’
2014
1909
Table 2 . Cardiac-Induced and Respiratory-Induced 3D Aortic Arch Translation (mm) at BA, LCCA, and LSA Branch Points BA (mm)
LCCA (mm)
LSA (mm)
0.2 ⫾ 1.8
0.2 ⫾ 1.3
Left-to-right translation Diastole-systole
Inspiration
0.3 ⫾ 1.8
Expiration-inspiration
Expiration Diastole
0.7 ⫾ 1.2* 1.3 ⫾ 2.7
Systole
0.9 ⫾ 2.8
0.4 ⫾ 2.8
0.5 ⫾ 2.2
Inspiration
0.4 ⫾ 1.5
0.0 ⫾ 1.4
0.1 ⫾ 1.6
Expiration
0.8 ⫾ 1.2*
1.1 ⫾ 1.4*
0.6 ⫾ 1.7
Diastole Systole
3.9 ⫾ 4.6* 3.4 ⫾ 4.3*
3.3 ⫾ 4.0* 2.3 ⫾ 3.9*
2.8 ⫾ 3.7* 2.2 ⫾ 3.7*
Posterior-to-anterior translation Diastole-systole Expiration-inspiration
0.8 ⫾ 1.0* 0.6 ⫾ 2.9
1.0 ⫾ 0.9* 0.2 ⫾ 2.6
Inferior-to-superior translation Diastole-systole
Inspiration Expiration
0.7 ⫾ 1.7 0.6 ⫾ 2.3
0.3 ⫾ 1.7 1.0 ⫾ 2.5
0.8 ⫾ 1.3* 0.2 ⫾ 2.3
Expiration-inspiration
Diastole
2.6 ⫾ 3.9*
3.2 ⫾ 3.6*
3.4 ⫾ 3.7*
Systole Inspiration
2.8 ⫾ 3.8* 2.4 ⫾ 1.7
2.5 ⫾ 3.3* 2.2 ⫾ 1.7
4.0 ⫾ 3.4* 2.0 ⫾ 1.5
Diastole-systole
Expiration
2.5 ⫾ 1.7
2.9 ⫾ 1.8
2.8 ⫾ 1.5
Expiration-inspiration
Diastole Systole
6.7 ⫾ 4.4 6.6 ⫾ 3.9
6.3 ⫾ 4.2 5.6 ⫾ 3.5
6.1 ⫾ 3.9 6.4 ⫾ 3.1
3D translation
Note.–Values are mean ⫾ SD. The significance threshold (P o .05) was adjusted by Bonferroni-Holm correction for multiple comparisons. BA ¼ brachiocephalic artery, LCCA ¼ left common carotid artery, LSA ¼ left subclavian artery, 3D ¼ three-dimensional *Statistically significant translation from one physiologic mode to another. Table 3 . Branching Angle (degrees) of BA, LCCA, and LSA at Four Physiologic Modes with Branch Vessel Vectors Defined by Points 10, 20, and 30 mm Distal to the Branch Ostia
Branch vector 10 mm distal Inspiration
BA (mm)
LCCA (mm)
LSA (mm)
Diastole
58.9 ⫾ 18.7
52.8 ⫾ 15.4*
65.6 ⫾ 19.5*
Systole
58.6 ⫾ 20.6
51.8 ⫾ 15.2*
65.5 ⫾ 17.4*
Diastole Systole
60.4 ⫾ 21.2 60.0 ⫾ 18.8
55.1 ⫾ 17.5 54.7 ⫾ 14.3*
66.0 ⫾ 20.1 65.8 ⫾ 19.6*
Inspiration
Diastole Systole
56.9 ⫾ 20.5† 55.0 ⫾ 22.5
48.2 ⫾ 16.3*† 46.6 ⫾ 15.6*
63.1 ⫾ 19.5* 63.1 ⫾ 22.5*
Expiration
Diastole
59.6 ⫾ 26.9
50.3 ⫾ 18.8*
63.8 ⫾ 23.0*
Systole
57.5 ⫾ 23.0
50.3 ⫾ 14.9
63.1 ⫾ 21.2
Diastole
59.9 ⫾ 25.7†
44.7 ⫾ 18.3*†
61.9 ⫾ 23.4*
Systole Diastole
†
58.0 ⫾ 26.7 63.2 ⫾ 31.6†
43.7 ⫾ 17.0*† 47.3 ⫾ 21.0*†
61.5 ⫾ 20.9* 62.8 ⫾ 23.8*
Systole
60.9 ⫾ 27.9†
47.4 ⫾ 18.5*†
62.2 ⫾ 22.6*
Expiration Branch vector 20 mm distal
Branch vector 30 mm distal Inspiration Expiration
Note.–Values are mean ⫾ SD. The significance threshold (P o .05) was adjusted by Bonferroni-Holm correction for multiple comparisons. No significant differences were found between BA and LSA. No significant differences were found between the physiologic modes or between branch vector lengths. BA ¼ brachiocephalic artery, LCCA ¼ left common carotid artery, LSA ¼ left subclavian artery. *Significant difference between LCCA and LSA. † Significant difference between BA and LCCA.
Limitations This study has limitations that could be addressed in future work. First, the small sample size and heterogeneous cohort limit the power of statistical analysis. In
addition, the repaired ascending aorta in some of the patients may exaggerate aortic arch motion because of the shorter distance from the heart to the arch and the greater transmission of cardiac motion to the arch because of the
1910
’
Aortic Arch Geometries and Deformations
Suh et al
’
JVIR
Table 4 . Differences in Branching Angle (degrees) of BA, LCCA, and LSA between Four Physiologic Modes with Branch Vessel Vectors Defined by Points 10, 20, and 30 mm Distal to the Branch Ostia BA (mm)
LCCA (mm)
LSA (mm)
Inspiration
0.3 ⫾ 4.2
1.0 ⫾ 4.3
0.1 ⫾ 4.8
Expiration
0.5 ⫾ 6.7
0.3 ⫾ 7.8
0.2 ⫾ 7.0
Diastole Systole
1.6 ⫾ 7.9 1.4 ⫾ 5.0
2.3 ⫾ 4.7 3.0 ⫾ 5.6
0.5 ⫾ 6.3 0.3 ⫾ 7.1
Diastole-systole
Inspiration Expiration
1.9 ⫾ 3.6 2.1 ⫾ 7.7
1.6 ⫾ 3.0 0.0 ⫾ 7.9
0.0 ⫾ 4.1 0.7 ⫾ 5.7
Expiration-inspiration
Diastole
2.7 ⫾ 9.0
2.1 ⫾ 4.6
0.7 ⫾ 5.9
Systole
2.5 ⫾ 5.1
3.7 ⫾ 6.0
0.0 ⫾ 6.8
Diastole-systole
Inspiration
1.9 ⫾ 3.3
1.0 ⫾ 2.6
0.4 ⫾ 3.9
Expiration-inspiration
Expiration Diastole
2.4 ⫾ 7.2 3.3 ⫾ 8.4
0.1 ⫾ 7.1 2.6 ⫾ 5.4
0.6 ⫾ 5.5 0.9 ⫾ 5.8
Systole
2.9 ⫾ 5.4
3.7 ⫾ 6.6
0.7 ⫾ 7.3
Branch vector 10 mm distal Diastole-systole Expiration-inspiration Branch vector 20 mm distal
Branch vector 30 mm distal
Note.–Values are mean ⫾ SD. The significance threshold (P o .05) was adjusted by Bonferroni-Holm correction for multiple comparisons. No significant differences were found between the physiologic modes or between vessels. BA ¼ brachiocephalic artery, LCCA ¼ left common carotid artery, LSA ¼ left subclavian artery.
stiffness of the surgical graft. Second, motion artifacts secondary to fast systolic motion, irregular heart rate, and muscle movement during breath hold cause image blurring that affects the precision of lumen segmentation. Third, although we analyzed only snapshots at end systole, diastole, inspiration, and expiration, a full analysis of cardiac-resolved and respiratory-resolved images could be done to understand the progression during a complete physiologic event (27). Finally, cardiac-induced and respiratory-induced deformations of the thoracic aorta were not quantified separately but could dramatically affect the dynamics of the aortic arch (28,29). In conclusion, this study quantified the 3D translations of the aortic arch, branching angles of the arch vessels, and their changes secondary to cardiac pulsation and respiration. The aortic arch translated significantly rightward from systole to diastole and posteriorly and superiorly from inspiration to expiration. Respiratoryinduced translation was significantly greater than cardiac-induced translation. Although branching angles of the BA, LCCA, and LSA were different from each other, none of them deformed significantly with cardiac or respiratory influences. The geometric and deformation data of the arch branch vessels may aid in treatment planning, implantation technique, and design and evaluation of branched thoracic aortic endografts. In addition, these quantitative methods can be used in other anatomies and to investigate changes of vessel geometry and compliance before and after endovascular repair.
ACKNOWLEDGMENT This work was supported by a research gift from Medtronic Inc. We thank Lior Molvin, Daisha Marsh, Christoph
Zorich, Monglan Duong, and Fatin Alkhadra for help with CT imaging and Riley Marangi for help with modeling. We also thank all the patients for their participation.
REFERENCES 1. Desai ND, Burtch K, Moser W, et al. Long-term comparison of thoracic endovascular aortic repair (TEVAR) to open surgery for the treatment of thoracic aortic aneurysms. J Thorac Cardiovasc Surg 2012; 144: 604–611. 2. Preventza O, Cervera R, Cooley DA, et al. Acute type I aortic dissection: traditional versus hybrid repair with antegrade stent delivery to the descending thoracic aorta. J Thorac Cardiovasc Surg 2014; 148: 119–125. 3. Tian DH, Wan B, Di Eusanio M, Black D, Yan TD. A systematic review and meta-analysis on the safety and efficacy of the frozen elephant trunk technique in aortic arch surgery. Ann Cardiothorac Surg 2013; 2:581–591. 4. Bonser RS, Ranasinghe AM, Loubani M, et al. Evidence, lack of evidence, controversy, and debate in the provision and performance of the surgery of acute type A aortic dissection. J Am Coll Cardiol 2011; 58: 2455–2474. 5. Lopera J, Patino JH, Urbina C, et al. Endovascular treatment of complicated type-B aortic dissection with stent-grafts: midterm results. J Vasc Interv Radiol 2003; 14:195–203. 6. Lee M, Lee DY, Kim MD, et al. Outcomes of endovascular management for complicated chronic type B aortic dissection: effect of the extent of stent graft coverage and anatomic properties of aortic dissection. J Vasc Interv Radiol 2013; 24:1451–1460. 7. Moon MC, Morales JP, Greenberg RK. The aortic arch and ascending aorta: are they within the endovascular realm? Semin Vasc Surg 2007; 20:97–107. 8. Chuter TAM, Hiramoto JS, Chang C, et al. Branched stent-grafts: will these become the new standard? J Vasc Interv Radiol 2008; 19:S57–S62. 9. Greenberg RK, O’Neill S, Walker E, et al. Endovascular repair of thoracic aortic lesions with the Zenith TX1 and TX2 thoracic grafts: intermediateterm results. J Vasc Surg 2005; 41:589–596. 10. Nordon IM, Hinchliffe RJ, Morgan R, Loftus IM, Jahangiri M, Thomson MM. Progress in endovascular management of type A dissection. Eur J Vasc Endovasc Surg 2012; 44:406–410. 11. Bavaria JE, Appoo JJ, Makaroun MS, Verter J, Yu Z-F, Mitchell RS. Endovascular stent grafting versus open surgical repair of descending thoracic aortic aneurysms in low-risk patients: a multicenter comparative trial. J Thorac Cardiovasc Surg 2007; 133:369–377.
Volume 25
’
Number 12
’
December
’
2014
12. Cheng D, Martin J, Shennib H, et al. Endovascular aortic repair versus open surgical repair for descending thoracic aortic disease: a systematic review and meta-analysis of comparative studies. J Am Coll Cardiol 2010; 55:986–1001. 13. Weidman JM, Desai M, Iftekhar A, et al. Engineering goals for future thoracic endograft—how can we make them more effective? Prog Cardiovasc Disease 2013; 56:92–102. 14. Zarins CK, Taylor CA. Endovascular device design in the future: transformation from trial and error to computational design. J Endovasc Ther 2009; 16:I12–I21. 15. Silverman PM, Roberts S, Tefft MC, et al. Helical CT of the liver: clinical application of an automated computer technique, SmartPrep, for obtaining images with optimal contrast enhancement. AJR Am J Roentgenol 1995; 165:73–78. 16. Fleischmann D. How to design injection protocols for multiple detector-row CT angiography (MDCTA). Eur Radiol Suppl 2005; 15: E60–E65. 17. Wang KC. Level Set Methods for Computational Prototyping with Application to Hemodynamic Modeling. Ph.D. Thesis. Stanford, CA: Stanford University Press; 2001. 18. Wilson N, Wang K, Dutton RW, Taylor CA. A software framework for creating patient specific geometric models from medical imaging data for simulation based medical planning of vascular surgery. Lect Notes Comput Sci 2001; 2208:449–456. 19. Choi G, Cheng CP, Wilson NM, Taylor CA. Methods for quantifying three-dimensional deformation of arteries due to pulsatile and nonpulsatile forces: implications for the design of stents and stent grafts. Ann Biomed Eng 2009; 37:14–33. 20. Holm A. A simple sequentially rejective multiple test procedure. Scand J Stat 1979; 6:65–70.
1911
21. Schepens MA. Aortic arch replacement: the conventional ‘elephant trunk’ technique. Multimed Man Cardiothorac Surg 2007; 2007(102): mmcts.2006.002006. 22. Weber TF, Ganten MK, Böckler D, Geisbüsch P, Kauzor HU, von TenggKobligk H. Heartbeat-related displacement of the thoracic aorta in patients with chronic aortic dissection type B: quantification by dynamic CTA. Eur J Radiol 2009; 72:483–488. 23. Goel VR, Greenberg RK, Greenberg DP. Mathematical analysis of DICOM CT datasets: can endograft sizing be automated for complex anatomy? J Vasc Surg 2008; 47:1306–1312. 24. Scott AD, Keegan J, Firmin DN. Motion in cardiovascular MR imaging. Radiology 2009; 250:331–351. 25. Weber TF, Tetzlaff R, Rengier F, et al. Respiratory displacement of the thoracic aorta: physiological phenomenon with potential implications for thoracic endovascular repair. Cardiovasc Intervent Radiol 2009; 32: 658–665. 26. Maxim PG, Loo BW Jr, Shirazi H, Thorndyke B, Luxton G, Le QT. Quantification of motion of different thoracic locations using fourdimensional computed tomography: implications for radiotherapy planning. Int J Radiat Oncol Biol Phys 2007; 69:1395–1401. 27. Weber TF, Muller T, Biesdorf A, et al. True four-dimensional analysis of thoracic aortic displacement and distension using model-based segmentation of computed tomography angiography. Int J Cardiovasc Imaging 2014; 30:185–194. 28. Morrison TM, Choi G, Zarins CK, Taylor CA. Circumferential and longitudinal cyclic strain of the human thoracic aorta: age-related changes. J Vasc Surg 2009; 49:1029–1036. 29. Carrascosa P, Capunay C, Deviggiano A, et al. Thoracic aorta cardiaccycle related dynamic changes assessed with a 256-slice CT scanner. Cardiovasc Diagn Ther 2013; 3:125–128.
