- Email: [email protected]
In vitro validation of flow measurements in an aortic nitinol stent graft by velocity-encoded MRI
European Journal of Radiology 80 (2011) 163–167
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
In vitro validation of flow measurements in an aortic nitinol stent graft by velocity-encoded MRI Fabian Rengier a,b,∗,2 , Michael Delles c,3 , Tim Frederik Weber a,1 , Dittmar Böckler d,4 , Sebastian Ley a,1 , Hans-Ulrich Kauczor a,1 , Hendrik von Tengg-Kobligk a,b,1,2 a
University Hospital Heidelberg, Department of Diagnostic and Interventional Radiology, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany German Cancer Research Center (dkfz) Heidelberg, Department of Radiology, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany c Karlsruhe Institute of Technology (KIT), Department of Informatics, Institute for Anthropomatics, Adenauerring 2, 76131 Karlsruhe, Germany d University Hospital Heidelberg, Department of Vascular and Endovascular Surgery, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany b
a r t i c l e
i n f o
Article history: Received 1 August 2010 Accepted 26 August 2010 Keywords: Magnetic resonance imaging Blood flow velocity Aorta Stent graft Endovascular
a b s t r a c t Purpose: To validate flow measurements within an aortic nickel–titanium (nitinol) stent graft using velocity-encoded cine magnetic resonance imaging (VEC MRI) and to assess intraobserver agreement of repeated flow measurements. Materials and methods: An elastic tube phantom mimicking the descending aorta was developed with the possibility to insert an aortic nitinol stent graft. Different flow patterns (constant, sinusoidal and pulsatile aortic flow) were applied by a gear pump. A two-dimensional phase-contrast sequence was used to acquire VEC perpendicular cross-sections at six equidistant levels along the phantom. Each acquisition was performed twice with and without stent graft, and each dataset was analysed twice by the same reader. The percental difference of the measured flow volume to the gold standard (pump setting) was defined as the parameter for accuracy. Furthermore, the intraobserver agreement was assessed. Results: Mean accuracy of flow volume measurements was −0.45 ± 1.63% without stent graft and −0.18 ± 1.45% with stent graft. Slightly lower accuracy was obtained for aortic flow both without (−2.31%) and with (−1.29%) stent graft. Accuracy was neither influenced by the measurement position nor by repeated acquisitions. There was significant intraobserver agreement with an intraclass correlation coefficient of 0.87 (without stent graft, p < 0.001) and 0.80 (with stent graft, p < 0.001). The coefficient of variance was 0.25% without stent graft and 0.28% with stent graft. Conclusion: This study demonstrated high accuracy and excellent intraobserver agreement of flow measurements within an aortic nitinol stent graft using VEC MRI. VEC MRI may give new insights into the haemodynamic consequences of endovascular aortic repair. © 2010 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Endovascular repair (EVR) of the thoracic and abdominal aorta has emerged as an accepted treatment option in suitable cases of most adult aortic diseases [1,2]. Pre-operative assessment
∗ Corresponding author at: University Hospital Heidelberg, Department of Diagnostic and Interventional Radiology, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany. Tel.: +49 6221 56 6410; fax: +49 6221 56 5730. E-mail addresses: [email protected], [email protected] (F. Rengier), [email protected] (M. Delles), [email protected] (T.F. Weber), [email protected] (D. Böckler), [email protected] (S. Ley), [email protected] (H.-U. Kauczor), [email protected] (H. von Tengg-Kobligk). 1 Tel.: +49 6221 56 6410; fax: +49 6221 56 5730. 2 Tel.: +49 6221 42 2564; fax: +49 6221 42 2567. 3 Tel.: +49 721 608 8430; fax: +49 721 608 4077. 4 Tel.: +49 6221 56 6249; fax: +49 6221 56 5423. 0720-048X/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2010.08.038
currently relies mainly on computed tomography [3,4], especially because of the possibility for advanced three-dimensional planning [5,6]. The value of magnetic resonance imaging (MRI) in particular for follow-up has been recognized in recent years after the development of MRI compatible nickel–titanium (nitinol) stent grafts [3,4,7]. Previous studies showed that all relevant morphological information for follow-up after nitinol stent graft implantation can be obtained by MRI [8,9]. MRI also offers the additional advantage of providing information on blood flow by velocity-encoded cine (VEC) imaging [10,11]. Blood flow alterations after stent graft implantation might be important for long-term outcome after EVR [12–14]. So far, haemodynamics after EVR have been investigated by computational fluid dynamics, indicating that haemodynamic changes after EVR may increase drag forces on the stent graft [12,13]. Increased drag forces in turn may lead to stent graft migration. Studies of computational fluid dynamics also found areas of reduced velocity within the stent
164
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167
graft that might be associated with thrombus formation [14,15]. Those computational studies have not yet been validated by in vivo measurements though. VEC MRI could be used for validation if it provided accurate flow measurements within aortic stent grafts. The knowledge gain on blood flow alterations within aortic stent grafts could then be used to identify risk factors for later stent graft failure as well as to improve stent graft design. The feasibility of VEC MRI in nitinol stent grafts has only been demonstrated for pulmonary and peripheral arteries [16–18]. However, the accuracy of VEC MRI in an aortic nitinol stent graft has not yet been investigated to our knowledge. Since aortic flow exhibits very high systolic velocities and flow volumes it may be an especially challenging environment. The purpose of this study was to validate flow measurements within an aortic nitinol stent graft using VEC MRI in an aortic flow phantom and to assess intraobserver agreement, i.e. reproducibility of flow measurements by repeated analysis. Further objective was to determine whether repeated acquisition, flow waveform and position of measurement in relation to the stent graft have a significant impact on the accuracy of flow measurements. Baseline values were obtained by flow measurements without stent graft using identical settings. 2. Materials and methods 2.1. Phantom setup
Fig. 1. Adjusted pump setting to mimic the aortic target waveform. The additional late-systolic peak is probably caused by a wave reflection within the phantom. Note that the shape of the flow waveform does not influence the average flow over the cycle and thus the values calculated in this study.
(time steps) was 15 for constant flow, 60 for sinusoidal flow and 128 for aortic flow. Cross-sections perpendicular to the elastic tube were acquired at six levels, all at 5 cm distances to each other with level 1 being located 5 cm distally to the tube entry. All levels could be exactly located on MRI using an in-house developed threedimensional gadolinium-based marking system. Each acquisition was performed twice. For each flow waveform, setups without and with stent graft were measured after each other before continuing with the next waveform.
A straight elastic tube with an average inner diameter of 27.5 mm and a length of 40 cm mimicked the descending aorta. A straight rigid tube of sufficient length was integrated upstream to ensure full establishment of flow profiles. An adjustable valve was located downstream to control the resistance of the flow circuit. A mixture of 40% glycerol and 60% distilled water was used as bloodmimicking fluid because of its viscosity, density and relaxation times similar to blood [16]. An MRI compatible, computer controlled gear pump (CardioFlow 5000 MR, Shelley Medical Imaging Technologies, London, Ontario, Canada) generated all flow waveforms. For the setup with stent graft, a thoracic aortic nitinol stent graft (Gore TAG, W.L. Gore & Associates, Flagstaff, AZ, USA) with a diameter of 31 mm and a length of 15 cm was inserted 10 cm distally to the entry of the elastic tube.
2.4. Image analysis
2.2. Flow
2.5. Statistical analysis
Three different flow waveforms were used with the following pump settings: constant flow at flow rates of 50, 75 and 100 ml/s, all with a measured cycle length of 100 ms; sinusoidal flow with an average flow rate of 100 ml/s, a flow range from 0 to 200 ml/s and a cycle length of 420 ms; pulsatile aortic flow with an average flow rate of 51.1 ml/s, a maximum flow of 300 ml/s and a cycle length of 1024 ms (Fig. 1). The manufacturer of the pump guarantees an accuracy of ±3% [19].
Kolmogorov–Smirnov test showed normal distribution of the variables. Data is given as mean ± SD. Standard two-sided t-test for paired samples was used to test for significant differences in the accuracy between measurements with and without stent graft. Additionally, Bland–Altman plots were computed [21]. The limit of agreement was calculated as mean ± 1.96 × SD. For statistical analysis the six acquisition levels were subsumed to three measurement positions (Fig. 3): position A (off-centre/before and after stent graft), including levels 1 and 6; position B (middle/proximal and distal ends of stent graft), including levels 2 and 5; position C (centre/within stent graft), including levels 3 and 4. Univariate analysis of variance was applied to test the impact of repeated acquisition, flow waveform and measurement position on the accuracy. In case of statistical significance the differences for the subgroups of these factors were assessed using post hoc test with Bonferroni correction. Intraobserver agreement was assessed by determining the Shrout and Fleiss intraclass correlation coefficient for the relative values [22] and the coefficient of variance. The coefficient of variance was determined as the SD of the two analyses divided by their mean. Two-sided t-test for paired samples was used to test for significant differences between the coefficients of variance with and without stent graft.
2.3. Image acquisition A 1.5 T clinical MR scanner (Magnetom Avanto, Siemens, Erlangen, Germany) was used for image acquisition. The pump was equipped with a mechanism for generating an ECG trigger signal permitting synchronization of the data acquisition to the flow waveform. A standard two-dimensional, through-plane VEC phase-contrast sequence supplied by the manufacturer was applied with retrospective ECG gating and the following imaging parameters: Field of View, 320 mm × 240 mm; spatial resolution, 1.25 mm × 1.25 mm; slice thickness, 6 mm; repetition/echo time (TR/TE), 13.30/3.39 ms; 2 averages; flip angle, 30◦ ; velocity encoding gradient (VENC), 150 cm/s. The number of calculated phases
Quantitative analysis of the acquired flow data was performed with dedicated postprocessing software [20]. A region of interest was placed on the phase image and adjusted for each time step. Its final position was cross-checked with the magnitude image (Fig. 2). The average flow [ml/s] over the pump cycle was provided by the software. Analysis of all 1080 images for constant flow, 1440 images for sinusoidal flow and 3072 images for aortic flow were performed twice by the same reader. The difference between the measured flow and the gold standard (pump setting) was calculated. Furthermore, a relative value was computed as the calculated flow difference divided by the pump setting. The mean of the two relative values was defined as the parameter for accuracy.
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167
165
Fig. 2. (a) Representative cross-sectional magnitude image through the stent graft with segmentation of the lumen. (b) Phase image with colour-coded velocities shows laminar flow profile within the stent graft.
A p-value of ≤0.05 was considered to represent statistical significance. All analyses were performed with SPSS Version 16.0 (SPSS Inc., Chicago, IL, USA). 3. Results Small susceptibility artefacts were visible on the magnitude image around the nitinol stent graft struts (Fig. 2a). However, those artefacts did not cause major image distortion and the single struts could still be clearly distinguished. No artefacts could be detected on the phase image within the region of interest (Fig. 2b).
Fig. 4. The accuracy of flow measurements without (a) and with (b) stent graft is plotted separately for each flow waveform. The straight lines represent mean difference, the dotted lines limits of agreement (for values see text).
Mean measured flow over all levels in the setup without stent graft was 50.0 ± 0.4 ml/s, 74.9 ± 0.4 ml/s and 100.0 ± 0.8 ml/s for constant flow, 100.2 ± 2.1 ml/s for sinusoidal flow, and 49.9 ± 1.0 ml/s for aortic flow. Mean measured flow over all levels in the setup with stent graft was 50.3 ± 0.4 ml/s, 74.6 ± 0.5 ml/s and 100.1 ± 0.7 ml/s for constant flow, 100.3 ± 1.5 ml/s for sinusoidal flow, and 50.4 ± 1.1 ml/s for aortic flow. 3.1. Accuracy Table 1 summarizes the accuracy values. Bland–Altman plots (Fig. 4) show the mean accuracy of −0.45% (limits of agreement −3.64 to 2.74%) for flow measurements without stent and −0.18% (limits of agreement −3.02 to 2.66%) for flow measurements with stent. There was no statistically significant difference between the accuracy with stent graft compared to the accuracy without stent Table 1 Accuracy of velocity-encoded MRI in the setup without and with stent graft.
Fig. 3. Diagram of the six levels for image acquisition and the three positions for statistical analysis in the setup with stent graft: (A) before and after stent graft; (B) proximal and distal ends of stent graft; (C) within stent graft.
Flow waveform
Without stent graft
With stent graft
Constant Sinusoidal Aortic
−0.04 ± 0.69% 0.15 ± 2.11% −2.31 ± 1.89%
0.02 ± 0.89% 0.32 ± 1.55% −1.29 ± 2.28%
166
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167
graft (p = 0.2). Repeated acquisition and measurement position did not have a significant impact on accuracy for both the setup without (p = 0.98/0.52) and with stent graft (p = 0.95/0.06). Flow waveform had a significant influence on the accuracy for both the setup without (p < 0.001) and with stent graft (p = 0.009). Post hoc Bonferroni revealed that the accuracy for the aortic flow was significantly lower compared to the accuracy for the constant flow (p < 0.001 without, and p = 0.016 with stent graft) as well as for the sinusoidal flow (p < 0.001 without, and p = 0.016 with stent graft). There was no significant difference between the accuracies for constant and sinusoidal flow (p = 1.0 for both setups). 3.2. Intraobserver agreement There was significant intraobserver agreement with an intraclass correlation coefficient of 0.87 for the setup without stent graft (p < 0.001) and of 0.80 for the setup with stent graft (p < 0.001). The limits of agreement were −1.72% to 1.74% without stent graft and −2.04% to 1.91% with stent graft. Mean coefficients of variance were 0.25% for the setup without stent graft and 0.28% for the setup with stent graft without a statistically significant difference between the two setups (p = 0.62). 4. Discussion This study showed that flow measurements in an aortic nitinol stent graft obtained by VEC MRI were highly accurate. Accuracy and intraobserver agreement of VEC MRI were equally excellent for measurements with and without the nitinol stent graft. The accuracy was not influenced by the measurement position in relation to the stent graft. Furthermore, there was no error associated with repeated acquisitions. However, VEC MRI slightly underestimated aortic flow by approximately 1–2%, and this tendency was present with and without the stent graft. To our knowledge, this is the first study to demonstrate the accuracy of VEC MRI in an aortic nitinol stent graft. It has to be noted that only the average flow volume can be used for validating in-stent flow measurements, because other flow parameters, e.g. velocity, may be altered by the presence of a stent graft. In the presented phantom setup, measurement errors depended on the observer, the scanner and the sequence, as well as the gold standard flow. Consequently, the reported errors could be at least partially due to the potential error of the gold standard flow (±3%). The importance of VEC MRI with regards to EVR needs further investigation [23,24]. However, being non-invasive and radiationfree, VEC MRI could be effectively used in the follow-up of EVR complementing existing MRI protocols for surveillance after EVR [25]. The presented study showed that the same average flow measurements are obtained after stent graft insertion as well as in repeated acquisitions. This implies that changes of haemodynamics after EVR as well as during a longer follow-up period could be accurately detected by VEC MRI. Previous studies presented the accuracy of VEC MRI for measuring aortic flow [26–28]. In the present study we observed a tendency of VEC MRI to slightly underestimate aortic flow which had been indicated by previous studies [26,28]. Our study data shows that the presence of an aortic nitinol stent graft does not influence the accuracy of VEC MRI and the tendency for slight underestimation. This is in accordance with studies investigating VEC MRI in nitinol stent grafts for pulmonary and peripheral arteries [16–18]. This study has two major clinical implications. First, VEC MRI can be used to accurately determine haemodynamic consequences of EVR and to validate computational simulations. This includes parameters derived from VEC MRI like pulse wave velocity, wall
shear stress and pressure gradients, promising new insights into the effects of EVR on the diseased aorta. Second, the role of MRI in the follow-up after nitinol stent graft implantation may further be strengthened by providing accurate information on both morphology and function. A limitation of the presented study might be that it did not include different nitinol stent grafts. However, other nitinol stent grafts can be expected to have similar degrees of artefacts [8]. The findings of our investigation and previous studies suggest that nitinol stent grafts do not cause any major artefact that could significantly influence flow measurements [16,18]. It has to be noted that stent grafts fabricated with alloys others than nitinol may exhibit significantly greater artefacts [8,18]. Another limitation might be that we did not perform an in vivo analysis. However, the purpose of this study was to assess the accuracy in relation to an objective gold standard, i.e. the pump setting, which cannot be achieved in vivo. Additionally, our phantom setup mimicked the in vivo situation including an elastic tube, pulsatile aortic flow and blood-mimicking fluid. In conclusion, this study demonstrated high accuracy and excellent intraobserver agreement of flow measurements within an aortic nitinol stent graft using VEC phase-contrast MRI. VEC MRI may give new insights into the haemodynamic consequences of EVR and can be used to validate computational simulations. Clinical studies in patients undergoing EVR with long-term follow-up are warranted to investigate the clinical relevance of VEC MRI in EVR patients and to describe potential haemodynamic risk factors for stent graft failure. Acknowledgements Fabian Rengier received a grant from the German Research Foundation within the “Research training group 1126: Intelligent Surgery-Development of new computer-based methods for the future workplace in surgery”. References [1] 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(10):986–1001. [2] Katzen BT, Dake MD, MacLean AA, Wang DS. Endovascular repair of abdominal and thoracic aortic aneurysms. Circulation 2005;112(11):1663–75. [3] Ueda T, Fleischmann D, Rubin GD, Dake MD, Sze DY. Imaging of the thoracic aorta before and after stent-graft repair of aneurysms and dissections. Semin Thorac Cardiovasc Surg 2008;20(4):348–57. [4] von Tengg-Kobligk H, Weber TF, Rengier F, et al. Imaging modalities for the thoracic aorta. J Cardiovasc Surg (Torino) 2008;49(4):429–47. [5] Lell MM, Anders K, Uder M, et al. New techniques in CT angiography. Radiographics 2006;26(Suppl. 1):S45–62. [6] Rengier F, Weber TF, Giesel FL, Böckler D, Kauczor HU, von Tengg-Kobligk H. Centerline analysis of aortic CT angiographic examinations: benefits and limitations. AJR Am J Roentgenol 2009;192(5):W255–63. [7] Stavropoulos SW, Charagundla SR. Imaging techniques for detection and management of endoleaks after endovascular aortic aneurysm repair. Radiology 2007;243(3):641–55. [8] van der Laan MJ, Bartels LW, Bakker CJG, Viergever MA, Blankensteijn JD. Suitability of 7 aortic stent-graft models for MRI-based surveillance. J Endovasc Ther 2004;11(4):366–71. [9] Merkle EM, Klein S, Wisianowsky C, et al. Magnetic resonance imaging versus multislice computed tomography of thoracic aortic endografts. J Endovasc Ther 2002;9(Suppl. 2):I12–3. [10] Pelc NJ, Herfkens RJ, Shimakawa A, Enzmann DR. Phase contrast cine magnetic resonance imaging. Magn Reson Q 1991;7(4):229–54. [11] Gatehouse PD, Keegan J, Crowe LA, et al. Applications of phase-contrast flow and velocity imaging in cardiovascular MRI. Eur Radiol 2005;15(10):2172–84. [12] Fung GSK, Lam SK, Cheng SWK, Chow KW. On stent-graft models in thoracic aortic endovascular repair: a computational investigation of the hemodynamic factors. Comput Biol Med 2008;38(4):484–9. [13] Frauenfelder T, Lotfey M, Boehm T, Wildermuth S. Computational fluid dynamics: hemodynamic changes in abdominal aortic aneurysm after stent-graft implantation. Cardiovasc Intervent Radiol 2006;29(4):613–23. [14] Chong CK, How TV. Flow patterns in an endovascular stent-graft for abdominal aortic aneurysm repair. J Biomech 2004;37(1):89–97.
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167 [15] Chong CK, How TV, Harris PL. Flow visualization in a model of a bifurcated stent-graft. J Endovasc Ther 2005;12(4):435–45. [16] Walsh EG, Holton AD, Brott BC, Venugopalan R, Anayiotos AS. Magnetic resonance phase velocity mapping through NiTi stents in a flow phantom model. J Magn Reson Imaging 2005;21(1):59–65. [17] van Holten J, Kunz P, Mulder PGH, Pattynama PMT, Lamb HJ, van Dijk LC. MRvelocity mapping in vascular stents to assess peak systolic velocity. In vitro comparison of various stent designs made of stainless steel and nitinol. MAGMA 2002;15(1–3):52–7. [18] Kuehne T, Saeed M, Moore P, et al. Influence of blood-pool contrast media on MR imaging and flow measurements in the presence of pulmonary arterial stents in swine. Radiology 2002;223(2):439–45. [19] CardioFlow 5000 MR Computer-controlled Flow Pump System. Ontario, Canada: Shelley Medical Imaging Technologies London. Available via http://www.simutec.com/Media/pumps/Brochure-CardioFlow 5000MR.pdf [accessed 11 June 2010]. [20] Unterhinninghofen R, Ley S, Ley-Zaporozhan J, et al. Concepts for visualization of multidirectional phase-contrast MRI of the heart and large thoracic vessels. Acad Radiol 2008;15(3):361–9. [21] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1(8476):307–10.
167
[22] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86(2):420–8. [23] Dall’Armellina E, Hamilton CA, Hundley WG. Assessment of blood flow and valvular heart disease using phase-contrast cardiovascular magnetic resonance. Echocardiography 2007;24(2):207–16. [24] von Tengg-Kobligk H, Weber T, Ley S, et al. State-of-the-art aortic imaging. Gefässchirurgie 2009;14(2):143–57. [25] Schwope RB, Alper HJ, Talenfeld AD, Cohen EI, Lookstein RA. MR angiography for patient surveillance after endovascular repair of abdominal aortic aneurysms. AJR Am J Roentgenol 2007;188(4):W334–40. [26] Ley S, Unterhinninghofen R, Ley-Zaporozhan J, Schenk J, Kauczor HU, Szabo G. Validation of magnetic resonance phase-contrast flow measurements in the main pulmonary artery and aorta using perivascular ultrasound in a large animal model. Invest Radiol 2008;43(6):421–6. [27] Powell AJ, Maier SE, Chung T, Geva T. Phase-velocity cine magnetic resonance imaging measurement of pulsatile blood flow in children and young adults: in vitro and in vivo validation. Pediatr Cardiol 2000;21(2):104–10. [28] Laffon E, Valli N, Latrabe V, Franconi JM, Barat JL, Laurent F. A validation of a flow quantification by MR phase mapping software. Eur J Radiol 1998;27(2):166–72.
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
In vitro validation of flow measurements in an aortic nitinol stent graft by velocity-encoded MRI Fabian Rengier a,b,∗,2 , Michael Delles c,3 , Tim Frederik Weber a,1 , Dittmar Böckler d,4 , Sebastian Ley a,1 , Hans-Ulrich Kauczor a,1 , Hendrik von Tengg-Kobligk a,b,1,2 a
University Hospital Heidelberg, Department of Diagnostic and Interventional Radiology, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany German Cancer Research Center (dkfz) Heidelberg, Department of Radiology, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany c Karlsruhe Institute of Technology (KIT), Department of Informatics, Institute for Anthropomatics, Adenauerring 2, 76131 Karlsruhe, Germany d University Hospital Heidelberg, Department of Vascular and Endovascular Surgery, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany b
a r t i c l e
i n f o
Article history: Received 1 August 2010 Accepted 26 August 2010 Keywords: Magnetic resonance imaging Blood flow velocity Aorta Stent graft Endovascular
a b s t r a c t Purpose: To validate flow measurements within an aortic nickel–titanium (nitinol) stent graft using velocity-encoded cine magnetic resonance imaging (VEC MRI) and to assess intraobserver agreement of repeated flow measurements. Materials and methods: An elastic tube phantom mimicking the descending aorta was developed with the possibility to insert an aortic nitinol stent graft. Different flow patterns (constant, sinusoidal and pulsatile aortic flow) were applied by a gear pump. A two-dimensional phase-contrast sequence was used to acquire VEC perpendicular cross-sections at six equidistant levels along the phantom. Each acquisition was performed twice with and without stent graft, and each dataset was analysed twice by the same reader. The percental difference of the measured flow volume to the gold standard (pump setting) was defined as the parameter for accuracy. Furthermore, the intraobserver agreement was assessed. Results: Mean accuracy of flow volume measurements was −0.45 ± 1.63% without stent graft and −0.18 ± 1.45% with stent graft. Slightly lower accuracy was obtained for aortic flow both without (−2.31%) and with (−1.29%) stent graft. Accuracy was neither influenced by the measurement position nor by repeated acquisitions. There was significant intraobserver agreement with an intraclass correlation coefficient of 0.87 (without stent graft, p < 0.001) and 0.80 (with stent graft, p < 0.001). The coefficient of variance was 0.25% without stent graft and 0.28% with stent graft. Conclusion: This study demonstrated high accuracy and excellent intraobserver agreement of flow measurements within an aortic nitinol stent graft using VEC MRI. VEC MRI may give new insights into the haemodynamic consequences of endovascular aortic repair. © 2010 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Endovascular repair (EVR) of the thoracic and abdominal aorta has emerged as an accepted treatment option in suitable cases of most adult aortic diseases [1,2]. Pre-operative assessment
∗ Corresponding author at: University Hospital Heidelberg, Department of Diagnostic and Interventional Radiology, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany. Tel.: +49 6221 56 6410; fax: +49 6221 56 5730. E-mail addresses: [email protected], [email protected] (F. Rengier), [email protected] (M. Delles), [email protected] (T.F. Weber), [email protected] (D. Böckler), [email protected] (S. Ley), [email protected] (H.-U. Kauczor), [email protected] (H. von Tengg-Kobligk). 1 Tel.: +49 6221 56 6410; fax: +49 6221 56 5730. 2 Tel.: +49 6221 42 2564; fax: +49 6221 42 2567. 3 Tel.: +49 721 608 8430; fax: +49 721 608 4077. 4 Tel.: +49 6221 56 6249; fax: +49 6221 56 5423. 0720-048X/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2010.08.038
currently relies mainly on computed tomography [3,4], especially because of the possibility for advanced three-dimensional planning [5,6]. The value of magnetic resonance imaging (MRI) in particular for follow-up has been recognized in recent years after the development of MRI compatible nickel–titanium (nitinol) stent grafts [3,4,7]. Previous studies showed that all relevant morphological information for follow-up after nitinol stent graft implantation can be obtained by MRI [8,9]. MRI also offers the additional advantage of providing information on blood flow by velocity-encoded cine (VEC) imaging [10,11]. Blood flow alterations after stent graft implantation might be important for long-term outcome after EVR [12–14]. So far, haemodynamics after EVR have been investigated by computational fluid dynamics, indicating that haemodynamic changes after EVR may increase drag forces on the stent graft [12,13]. Increased drag forces in turn may lead to stent graft migration. Studies of computational fluid dynamics also found areas of reduced velocity within the stent
164
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167
graft that might be associated with thrombus formation [14,15]. Those computational studies have not yet been validated by in vivo measurements though. VEC MRI could be used for validation if it provided accurate flow measurements within aortic stent grafts. The knowledge gain on blood flow alterations within aortic stent grafts could then be used to identify risk factors for later stent graft failure as well as to improve stent graft design. The feasibility of VEC MRI in nitinol stent grafts has only been demonstrated for pulmonary and peripheral arteries [16–18]. However, the accuracy of VEC MRI in an aortic nitinol stent graft has not yet been investigated to our knowledge. Since aortic flow exhibits very high systolic velocities and flow volumes it may be an especially challenging environment. The purpose of this study was to validate flow measurements within an aortic nitinol stent graft using VEC MRI in an aortic flow phantom and to assess intraobserver agreement, i.e. reproducibility of flow measurements by repeated analysis. Further objective was to determine whether repeated acquisition, flow waveform and position of measurement in relation to the stent graft have a significant impact on the accuracy of flow measurements. Baseline values were obtained by flow measurements without stent graft using identical settings. 2. Materials and methods 2.1. Phantom setup
Fig. 1. Adjusted pump setting to mimic the aortic target waveform. The additional late-systolic peak is probably caused by a wave reflection within the phantom. Note that the shape of the flow waveform does not influence the average flow over the cycle and thus the values calculated in this study.
(time steps) was 15 for constant flow, 60 for sinusoidal flow and 128 for aortic flow. Cross-sections perpendicular to the elastic tube were acquired at six levels, all at 5 cm distances to each other with level 1 being located 5 cm distally to the tube entry. All levels could be exactly located on MRI using an in-house developed threedimensional gadolinium-based marking system. Each acquisition was performed twice. For each flow waveform, setups without and with stent graft were measured after each other before continuing with the next waveform.
A straight elastic tube with an average inner diameter of 27.5 mm and a length of 40 cm mimicked the descending aorta. A straight rigid tube of sufficient length was integrated upstream to ensure full establishment of flow profiles. An adjustable valve was located downstream to control the resistance of the flow circuit. A mixture of 40% glycerol and 60% distilled water was used as bloodmimicking fluid because of its viscosity, density and relaxation times similar to blood [16]. An MRI compatible, computer controlled gear pump (CardioFlow 5000 MR, Shelley Medical Imaging Technologies, London, Ontario, Canada) generated all flow waveforms. For the setup with stent graft, a thoracic aortic nitinol stent graft (Gore TAG, W.L. Gore & Associates, Flagstaff, AZ, USA) with a diameter of 31 mm and a length of 15 cm was inserted 10 cm distally to the entry of the elastic tube.
2.4. Image analysis
2.2. Flow
2.5. Statistical analysis
Three different flow waveforms were used with the following pump settings: constant flow at flow rates of 50, 75 and 100 ml/s, all with a measured cycle length of 100 ms; sinusoidal flow with an average flow rate of 100 ml/s, a flow range from 0 to 200 ml/s and a cycle length of 420 ms; pulsatile aortic flow with an average flow rate of 51.1 ml/s, a maximum flow of 300 ml/s and a cycle length of 1024 ms (Fig. 1). The manufacturer of the pump guarantees an accuracy of ±3% [19].
Kolmogorov–Smirnov test showed normal distribution of the variables. Data is given as mean ± SD. Standard two-sided t-test for paired samples was used to test for significant differences in the accuracy between measurements with and without stent graft. Additionally, Bland–Altman plots were computed [21]. The limit of agreement was calculated as mean ± 1.96 × SD. For statistical analysis the six acquisition levels were subsumed to three measurement positions (Fig. 3): position A (off-centre/before and after stent graft), including levels 1 and 6; position B (middle/proximal and distal ends of stent graft), including levels 2 and 5; position C (centre/within stent graft), including levels 3 and 4. Univariate analysis of variance was applied to test the impact of repeated acquisition, flow waveform and measurement position on the accuracy. In case of statistical significance the differences for the subgroups of these factors were assessed using post hoc test with Bonferroni correction. Intraobserver agreement was assessed by determining the Shrout and Fleiss intraclass correlation coefficient for the relative values [22] and the coefficient of variance. The coefficient of variance was determined as the SD of the two analyses divided by their mean. Two-sided t-test for paired samples was used to test for significant differences between the coefficients of variance with and without stent graft.
2.3. Image acquisition A 1.5 T clinical MR scanner (Magnetom Avanto, Siemens, Erlangen, Germany) was used for image acquisition. The pump was equipped with a mechanism for generating an ECG trigger signal permitting synchronization of the data acquisition to the flow waveform. A standard two-dimensional, through-plane VEC phase-contrast sequence supplied by the manufacturer was applied with retrospective ECG gating and the following imaging parameters: Field of View, 320 mm × 240 mm; spatial resolution, 1.25 mm × 1.25 mm; slice thickness, 6 mm; repetition/echo time (TR/TE), 13.30/3.39 ms; 2 averages; flip angle, 30◦ ; velocity encoding gradient (VENC), 150 cm/s. The number of calculated phases
Quantitative analysis of the acquired flow data was performed with dedicated postprocessing software [20]. A region of interest was placed on the phase image and adjusted for each time step. Its final position was cross-checked with the magnitude image (Fig. 2). The average flow [ml/s] over the pump cycle was provided by the software. Analysis of all 1080 images for constant flow, 1440 images for sinusoidal flow and 3072 images for aortic flow were performed twice by the same reader. The difference between the measured flow and the gold standard (pump setting) was calculated. Furthermore, a relative value was computed as the calculated flow difference divided by the pump setting. The mean of the two relative values was defined as the parameter for accuracy.
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167
165
Fig. 2. (a) Representative cross-sectional magnitude image through the stent graft with segmentation of the lumen. (b) Phase image with colour-coded velocities shows laminar flow profile within the stent graft.
A p-value of ≤0.05 was considered to represent statistical significance. All analyses were performed with SPSS Version 16.0 (SPSS Inc., Chicago, IL, USA). 3. Results Small susceptibility artefacts were visible on the magnitude image around the nitinol stent graft struts (Fig. 2a). However, those artefacts did not cause major image distortion and the single struts could still be clearly distinguished. No artefacts could be detected on the phase image within the region of interest (Fig. 2b).
Fig. 4. The accuracy of flow measurements without (a) and with (b) stent graft is plotted separately for each flow waveform. The straight lines represent mean difference, the dotted lines limits of agreement (for values see text).
Mean measured flow over all levels in the setup without stent graft was 50.0 ± 0.4 ml/s, 74.9 ± 0.4 ml/s and 100.0 ± 0.8 ml/s for constant flow, 100.2 ± 2.1 ml/s for sinusoidal flow, and 49.9 ± 1.0 ml/s for aortic flow. Mean measured flow over all levels in the setup with stent graft was 50.3 ± 0.4 ml/s, 74.6 ± 0.5 ml/s and 100.1 ± 0.7 ml/s for constant flow, 100.3 ± 1.5 ml/s for sinusoidal flow, and 50.4 ± 1.1 ml/s for aortic flow. 3.1. Accuracy Table 1 summarizes the accuracy values. Bland–Altman plots (Fig. 4) show the mean accuracy of −0.45% (limits of agreement −3.64 to 2.74%) for flow measurements without stent and −0.18% (limits of agreement −3.02 to 2.66%) for flow measurements with stent. There was no statistically significant difference between the accuracy with stent graft compared to the accuracy without stent Table 1 Accuracy of velocity-encoded MRI in the setup without and with stent graft.
Fig. 3. Diagram of the six levels for image acquisition and the three positions for statistical analysis in the setup with stent graft: (A) before and after stent graft; (B) proximal and distal ends of stent graft; (C) within stent graft.
Flow waveform
Without stent graft
With stent graft
Constant Sinusoidal Aortic
−0.04 ± 0.69% 0.15 ± 2.11% −2.31 ± 1.89%
0.02 ± 0.89% 0.32 ± 1.55% −1.29 ± 2.28%
166
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167
graft (p = 0.2). Repeated acquisition and measurement position did not have a significant impact on accuracy for both the setup without (p = 0.98/0.52) and with stent graft (p = 0.95/0.06). Flow waveform had a significant influence on the accuracy for both the setup without (p < 0.001) and with stent graft (p = 0.009). Post hoc Bonferroni revealed that the accuracy for the aortic flow was significantly lower compared to the accuracy for the constant flow (p < 0.001 without, and p = 0.016 with stent graft) as well as for the sinusoidal flow (p < 0.001 without, and p = 0.016 with stent graft). There was no significant difference between the accuracies for constant and sinusoidal flow (p = 1.0 for both setups). 3.2. Intraobserver agreement There was significant intraobserver agreement with an intraclass correlation coefficient of 0.87 for the setup without stent graft (p < 0.001) and of 0.80 for the setup with stent graft (p < 0.001). The limits of agreement were −1.72% to 1.74% without stent graft and −2.04% to 1.91% with stent graft. Mean coefficients of variance were 0.25% for the setup without stent graft and 0.28% for the setup with stent graft without a statistically significant difference between the two setups (p = 0.62). 4. Discussion This study showed that flow measurements in an aortic nitinol stent graft obtained by VEC MRI were highly accurate. Accuracy and intraobserver agreement of VEC MRI were equally excellent for measurements with and without the nitinol stent graft. The accuracy was not influenced by the measurement position in relation to the stent graft. Furthermore, there was no error associated with repeated acquisitions. However, VEC MRI slightly underestimated aortic flow by approximately 1–2%, and this tendency was present with and without the stent graft. To our knowledge, this is the first study to demonstrate the accuracy of VEC MRI in an aortic nitinol stent graft. It has to be noted that only the average flow volume can be used for validating in-stent flow measurements, because other flow parameters, e.g. velocity, may be altered by the presence of a stent graft. In the presented phantom setup, measurement errors depended on the observer, the scanner and the sequence, as well as the gold standard flow. Consequently, the reported errors could be at least partially due to the potential error of the gold standard flow (±3%). The importance of VEC MRI with regards to EVR needs further investigation [23,24]. However, being non-invasive and radiationfree, VEC MRI could be effectively used in the follow-up of EVR complementing existing MRI protocols for surveillance after EVR [25]. The presented study showed that the same average flow measurements are obtained after stent graft insertion as well as in repeated acquisitions. This implies that changes of haemodynamics after EVR as well as during a longer follow-up period could be accurately detected by VEC MRI. Previous studies presented the accuracy of VEC MRI for measuring aortic flow [26–28]. In the present study we observed a tendency of VEC MRI to slightly underestimate aortic flow which had been indicated by previous studies [26,28]. Our study data shows that the presence of an aortic nitinol stent graft does not influence the accuracy of VEC MRI and the tendency for slight underestimation. This is in accordance with studies investigating VEC MRI in nitinol stent grafts for pulmonary and peripheral arteries [16–18]. This study has two major clinical implications. First, VEC MRI can be used to accurately determine haemodynamic consequences of EVR and to validate computational simulations. This includes parameters derived from VEC MRI like pulse wave velocity, wall
shear stress and pressure gradients, promising new insights into the effects of EVR on the diseased aorta. Second, the role of MRI in the follow-up after nitinol stent graft implantation may further be strengthened by providing accurate information on both morphology and function. A limitation of the presented study might be that it did not include different nitinol stent grafts. However, other nitinol stent grafts can be expected to have similar degrees of artefacts [8]. The findings of our investigation and previous studies suggest that nitinol stent grafts do not cause any major artefact that could significantly influence flow measurements [16,18]. It has to be noted that stent grafts fabricated with alloys others than nitinol may exhibit significantly greater artefacts [8,18]. Another limitation might be that we did not perform an in vivo analysis. However, the purpose of this study was to assess the accuracy in relation to an objective gold standard, i.e. the pump setting, which cannot be achieved in vivo. Additionally, our phantom setup mimicked the in vivo situation including an elastic tube, pulsatile aortic flow and blood-mimicking fluid. In conclusion, this study demonstrated high accuracy and excellent intraobserver agreement of flow measurements within an aortic nitinol stent graft using VEC phase-contrast MRI. VEC MRI may give new insights into the haemodynamic consequences of EVR and can be used to validate computational simulations. Clinical studies in patients undergoing EVR with long-term follow-up are warranted to investigate the clinical relevance of VEC MRI in EVR patients and to describe potential haemodynamic risk factors for stent graft failure. Acknowledgements Fabian Rengier received a grant from the German Research Foundation within the “Research training group 1126: Intelligent Surgery-Development of new computer-based methods for the future workplace in surgery”. References [1] 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(10):986–1001. [2] Katzen BT, Dake MD, MacLean AA, Wang DS. Endovascular repair of abdominal and thoracic aortic aneurysms. Circulation 2005;112(11):1663–75. [3] Ueda T, Fleischmann D, Rubin GD, Dake MD, Sze DY. Imaging of the thoracic aorta before and after stent-graft repair of aneurysms and dissections. Semin Thorac Cardiovasc Surg 2008;20(4):348–57. [4] von Tengg-Kobligk H, Weber TF, Rengier F, et al. Imaging modalities for the thoracic aorta. J Cardiovasc Surg (Torino) 2008;49(4):429–47. [5] Lell MM, Anders K, Uder M, et al. New techniques in CT angiography. Radiographics 2006;26(Suppl. 1):S45–62. [6] Rengier F, Weber TF, Giesel FL, Böckler D, Kauczor HU, von Tengg-Kobligk H. Centerline analysis of aortic CT angiographic examinations: benefits and limitations. AJR Am J Roentgenol 2009;192(5):W255–63. [7] Stavropoulos SW, Charagundla SR. Imaging techniques for detection and management of endoleaks after endovascular aortic aneurysm repair. Radiology 2007;243(3):641–55. [8] van der Laan MJ, Bartels LW, Bakker CJG, Viergever MA, Blankensteijn JD. Suitability of 7 aortic stent-graft models for MRI-based surveillance. J Endovasc Ther 2004;11(4):366–71. [9] Merkle EM, Klein S, Wisianowsky C, et al. Magnetic resonance imaging versus multislice computed tomography of thoracic aortic endografts. J Endovasc Ther 2002;9(Suppl. 2):I12–3. [10] Pelc NJ, Herfkens RJ, Shimakawa A, Enzmann DR. Phase contrast cine magnetic resonance imaging. Magn Reson Q 1991;7(4):229–54. [11] Gatehouse PD, Keegan J, Crowe LA, et al. Applications of phase-contrast flow and velocity imaging in cardiovascular MRI. Eur Radiol 2005;15(10):2172–84. [12] Fung GSK, Lam SK, Cheng SWK, Chow KW. On stent-graft models in thoracic aortic endovascular repair: a computational investigation of the hemodynamic factors. Comput Biol Med 2008;38(4):484–9. [13] Frauenfelder T, Lotfey M, Boehm T, Wildermuth S. Computational fluid dynamics: hemodynamic changes in abdominal aortic aneurysm after stent-graft implantation. Cardiovasc Intervent Radiol 2006;29(4):613–23. [14] Chong CK, How TV. Flow patterns in an endovascular stent-graft for abdominal aortic aneurysm repair. J Biomech 2004;37(1):89–97.
F. Rengier et al. / European Journal of Radiology 80 (2011) 163–167 [15] Chong CK, How TV, Harris PL. Flow visualization in a model of a bifurcated stent-graft. J Endovasc Ther 2005;12(4):435–45. [16] Walsh EG, Holton AD, Brott BC, Venugopalan R, Anayiotos AS. Magnetic resonance phase velocity mapping through NiTi stents in a flow phantom model. J Magn Reson Imaging 2005;21(1):59–65. [17] van Holten J, Kunz P, Mulder PGH, Pattynama PMT, Lamb HJ, van Dijk LC. MRvelocity mapping in vascular stents to assess peak systolic velocity. In vitro comparison of various stent designs made of stainless steel and nitinol. MAGMA 2002;15(1–3):52–7. [18] Kuehne T, Saeed M, Moore P, et al. Influence of blood-pool contrast media on MR imaging and flow measurements in the presence of pulmonary arterial stents in swine. Radiology 2002;223(2):439–45. [19] CardioFlow 5000 MR Computer-controlled Flow Pump System. Ontario, Canada: Shelley Medical Imaging Technologies London. Available via http://www.simutec.com/Media/pumps/Brochure-CardioFlow 5000MR.pdf [accessed 11 June 2010]. [20] Unterhinninghofen R, Ley S, Ley-Zaporozhan J, et al. Concepts for visualization of multidirectional phase-contrast MRI of the heart and large thoracic vessels. Acad Radiol 2008;15(3):361–9. [21] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1(8476):307–10.
167
[22] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86(2):420–8. [23] Dall’Armellina E, Hamilton CA, Hundley WG. Assessment of blood flow and valvular heart disease using phase-contrast cardiovascular magnetic resonance. Echocardiography 2007;24(2):207–16. [24] von Tengg-Kobligk H, Weber T, Ley S, et al. State-of-the-art aortic imaging. Gefässchirurgie 2009;14(2):143–57. [25] Schwope RB, Alper HJ, Talenfeld AD, Cohen EI, Lookstein RA. MR angiography for patient surveillance after endovascular repair of abdominal aortic aneurysms. AJR Am J Roentgenol 2007;188(4):W334–40. [26] Ley S, Unterhinninghofen R, Ley-Zaporozhan J, Schenk J, Kauczor HU, Szabo G. Validation of magnetic resonance phase-contrast flow measurements in the main pulmonary artery and aorta using perivascular ultrasound in a large animal model. Invest Radiol 2008;43(6):421–6. [27] Powell AJ, Maier SE, Chung T, Geva T. Phase-velocity cine magnetic resonance imaging measurement of pulsatile blood flow in children and young adults: in vitro and in vivo validation. Pediatr Cardiol 2000;21(2):104–10. [28] Laffon E, Valli N, Latrabe V, Franconi JM, Barat JL, Laurent F. A validation of a flow quantification by MR phase mapping software. Eur J Radiol 1998;27(2):166–72.