 Research
 Open access
 Published:
An adaptive optimal viewing angle determination algorithm for TEVAR operation
BMC Medical Imaging volume 21, Article number: 141 (2021)
Abstract
Background
The determination of the right xray angiography viewing angle is an important issue during the treatment of thoracic endovascular aortic repair (TEVAR). An inaccurate projection angle (manually determined today by the physicians according to their personal experience) may affect the placement of the stent and cause vascular occlusion or endoleak.
Methods
Based on the acquisition of a computed tomography angiography (CTA) image before TEVAR, an adaptive optimization algorithm is proposed to determine the optimal viewing angle of the angiogram automatically. This optimal view aims at avoiding any overlapping between the left common carotid artery and the left subclavian artery. Moreover, the proposed optimal procedure exploits the patientspecific morphology to adaptively reduce the potential foreshortening effect.
Results
Experimental results conducted on thirtyfive patients demonstrate that the optimal angiographic viewing angle based on the proposed method has no significant difference when compared with the expert practice (p = 0.0678).
Conclusion
We propose a method that utilizes the CTA image acquired before TEVAR to automatically calculate the optimal Carm angle. This method has the potential to assist surgeons during their interventional procedure by providing a shorter procedure time, less radiation exposure, and less contrast injection.
Background
Type B aortic dissection (TBAD) is an acute cardiovascular disease with high mortality and disability rates [1,2,3]. Medical treatment is routinely used in the clinic to treat the uncomplicated aortic dissection. However, for the complicated TBAD, the thoracic endovascular aortic repair (TEVAR) is an alternative and effective tool compared to open surgery [4,5,6]. Since the stent placement is mainly guided by digital subtraction angiography (DSA), an inaccurate viewing angle may lead to 3D vascular structures overlap in the projected 2D images. Therefore, the viewing angle of the Carm plays an important role in placing the stent grafts.
During the process of TEVAR, physicians manually determine the angiographic viewing angles according to their personal experiences in order to display the whole aortic arch while eliminating any overlap of branches. This task is subjective and may not only be timeconsuming for young physicians but also increase the patient’s intake of Xray and contrast agent. Furthermore, the Xray angiographic images with improper viewing angles lose much aortic topological information [7, 8], thus, affect the accuracy of the stent placement and lead to some later complications such as aortic rupture or retrograde dissection [9]. The current clinical imaging angle selection requires the participation of expert cardiologists, hence an additional burden in the clinical setting. Therefore, the automatic selection of angiography viewing angle is extremely important and clinically required.
The issue of automatic optimum viewing angle determination has attracted many researchers’ attention because of its major clinical implications. Dumay et al. [10] assumed that the interested vessel is a cylinder and selected two projection images of different angles. Then, the optimal viewing angle was obtained through the geometry relationship of corresponding vectors. Nevertheless, this method is only suitable for slightlycurved vessels and requires at least two different angles of projection images. It fails in case of large and extended curvature. Moreover, obtaining projection images at different angles increase the radiation dose. Considering the intake of Xray and contrast agent, a series of studies used CTA data to obtain the optimal angle in the coronary artery [11,12,13,14,15,16]. The vessel vector was represented by the line between two adjacent points on the centerline of the vessel of interest. Chen et al. [17] used a predefined threshold to calculate the minimum vascular projection foreshortening rate. The optimal angiographic angle was obtained by refining the overlapping rate. Considering that the aortic branches are much larger than the coronary arteries, the method applied in coronary arteries is not suitable for the aorta. The imaging angle automatically obtained through mathematically minimum methods does not fit the clinical scenario in the aorta. The optimal angle obtained by adding the constraint design of the realistic clinical scenario is more suitable for the clinic.
In order to solve the aforementioned drawbacks, we proposed an adaptive optimization method that automatically determines the angiography viewing angle for the TEVAR operation. The projection foreshortening rate (PFR) is used to obtain a full display of the aortic arch, and the projection overlapping rate (POR) is used to avoid the overlap of branches on the aortic arch. By combining an empirical regularization term, the optimal angle can be automatically obtained before TEVAR, which ensures doctors a better imaging angle to place the stent more accurately. The contributions of this work are summarized as follows:

1.
To our knowledge, our solution is the first attempt based on preoperative CTA images for providing an automatically and reliably optimal imaging angle view for TEVAR.

2.
The optimum viewing angle is obtained by optimizing the display of the aortic arch without overlapping branches instead of minimizing the POR and PFR.

3.
Combined with the clinical scenario of the aorta, the proposed method adaptively provides junior doctors with an optimal imaging angle preoperation.

4.
This method has the potential to optimize the planning time, improve the accuracy of the stent placement and reduce the radiation exposure during TEVAR.
Methods
Patients with aortic dissection before TEVAR usually have a CTA examination. Based on the CTA data, the aorta is segmented and the centerline is obtained. CTA data are then used to simulate angiograms. The amount of foreshortening and vessel overlap for the aortic arch can be calculated. Without the left common carotid artery (LCCA) and left subclavian artery (LSA) overlapped, an adaptive optimization approach is proposed to determine the viewing angle before the operation automatically. Figure 1 shows the flow chart of the proposed algorithm.
Patients and image acquisition
Our institutional review board approved this study. We focus on the TBAD patients who needed the interventional operation from August 2019 to December 2020: 63 patients in the acute phase underwent TEVAR identified from 107 type B aortic dissection patients. Detailed inhospital information is given in Table 1. Among these 63 patients, we excluded the patients with procedure unsuccess (n = 2), patients who died in the hospital (n = 3), patients submitted to the adjuvant procedure (n = 10), patients who lost DSA or CTA data before intervention (n = 10) and TEVAR covered left subclavian artery (n = 3) from the dataset. The final patient set consists of 35 patients. The basic characteristics of patients are summarized in Table 2. Each CT angiography dataset is acquired before stentgraft implantation via a dualsource CT scanner (SOMATOM Definition Flash, SIEMENS, Germany) and the intervention is performed under the DSA Artis Zee (SIEMENS, Germany). Two experts (cardiovascular radiologists) work together to manually acquire the viewing angle of the DSA during TEVAR for all the patients. During the CTA, the patient lies on his back on the table with his upper limbs raised vertically above the head and is placed in a supine position during the intervention with his arms down naturally at his sides.
Projection overlapping rate
The 2D projection principle in Xray imaging leads to the superimposition of all tissues and, in our case, may lead to an overlap of the branches connected to the aortic arch (i.e. LCCA and LSA). This overlap can affect the placement of the stent in patients with type B aortic dissection. The blue square in Fig. 2 shows an example of such overlap where it is difficult to distinguish the boundaries of each branch. Improper placement of stents covering the branches of the aortic arch can cause organ ischemia. The optimum viewing angle should therefore be estimated by minimizing the overlapping rate of the LCCA and LSA.
Let S_{LCCA}(θ) and S_{LSA}(θ) represent the LCCA and LSA segments in the projection image at angle θ respectively, the POR P_{overlap} can be calculated by counting the pixels in the projection image as:
where NP(A) denotes the number of pixels of A.
Projection foreshortening rate
In the 2D angiographic image, when the Xray direction is not perpendicular to the aortic arch plane, the projection of the length of the aorta in the twodimensional image is shorter than the length in 3D. In Fig. 3, we simulate the angiographic image by computing the Xray projections from different directions and measure the length of the aorta. It is estimated in image A to 26 cm, much shorter than the actual length (30 cm) in 3D. In image B, the aortic arch is fully displayed and does not have any foreshortening. Direction B can be almost perpendicular to the aortic arch plane. It is difficult to determine the best placement of the stent when the aortic arch overlap. Displaying the aortic arch as completely as possible in 2D angiography permits the physicians to observe the aortic structure information clearly, thus facilitate the accurate placement of the stent. The PFR P_{fs} of the aortic arch can be calculated by the ratio of the centerline length after projection under angle θ to the centerline length in CTA:
where \(\overrightarrow {P}\) represents the projection vector of xray, \(\overrightarrow {C} (t)\) is the centerline of the aorta in 3D, and T_{1}(θ) represents the total length of the centerline obtained at angle θ in 2D projection and T_{2} represents the total length of the centerline in 3D.
Standard algorithm
Minimizing the foreshortening rate and the overlapping rate can be expressed as solving the standard optimization problem:
However such approach has two drawbacks here: since the sum of POR and PFR is minimized, there is no guarantee that the optimum angle will suppress any overlapping between LCCA and LSA. In addition, the morphology of the aorta is not taken into account. Therefore it seems more meaningful to adopt a two steps method by looking for angles eliminating the superimpositions and then by selecting among them the optimal view satisfying the PFR P_{fs} criteria.
Adaptive algorithm
The detail of our algorithm to obtain the optimal view can be described as follows:
Step 1: The aorta in the CTA data is segmented and the centerline is obtained by the method reported in Ref. [18].
Step 2: Angiographic images are simulated for angle θ varying from 0 to 90 degrees of the Carm based on the 3D vascular structure.
Step 3: Calculate the POR between LCCA and LSA by Eq. (1), and denote θ_{nooverlap} the set of angles which lead to P_{overlap}(θ) = 0, that means, there is no overlap between LCCA and LSA.
Step 4: Calculate the PFR by Equation (2) for each given value of θ, and we choose the angle θ_{fsmin} which reaches the maximal value of foreshortening P_{fs} denoted hereafter by P_{fsmax}. If the angle θ_{fsmin} is too large, an adaptive strategy is adopted to meet the clinical requirement. To achieve this goal, we define:
Step 5: Calculate the intersection of θ_{nooverlap} and θ_{fs}, the minimal value of this interval is chosen as the optimal view angle θ_{optimal}.
Results
Qualitative and quantitative analyses are performed on the dataset described above and comparisons are carried out between the standard approach, the proposed adaptive method using the ground truth of Carm angles set by the cardiovascular radiologist during TEVAR.
Quantitative analysis
The statistical analysis is performed using Prism 7.0 (GraphPad Software Inc, San Diego, Calif). If the assumption of normality is met, statistically significant differences for the viewing angle result obtained from different algorithms are evaluated using the Ttest. However, if the assumption of normality is not met, a Mann–Whitney U test was performed. Values of P < 0.05 are considered statistically significant.
Figure 4a shows the optimal viewing angles obtained for 35 aortic dissection patients. The means and standard deviations are respectively 54.67 ± 1.645 (standard approach), 44.55 ± 0.5014 (adaptive method) and 45.79 ± 0.4474 (experts). The differences between the standard algorithm and experts are significant (p < 0.0001), no significant difference is observed between the experts and the adaptive results (p = 0.0678).
Figure 4b summarizes these results. The red line depicts the absolute angle difference between experts and the standard minimization approach and the green line is the absolute difference between our proposed algorithm and expert settings. It can be seen that the adaptive viewing angle optimization has less error than the standard one. The distributions of the angle differences between the three frames are presented in Fig. 4c and d, respectively. While in the former case (4C) only twenty percent of differences are less than three degrees, when looking at the adaptive versus experts angle setting, the differences are less than five degrees, and most of them less than three degrees.
Qualitative analysis
In order to evaluate the performance of the proposed method, the clinical CTA is used to simulate a 2D angiographic imaging system. Figure 5 shows the simulated angiographic view by experts (first row) used during the intervention, the standard (second row) and adaptive algorithm (third row) results obtained on four patients. In Fig. 5C1 and C2 display the absolute difference image between the ground truth projection and the projections obtained through the standard and adaptive algorithms respectively. A color image coding is adopted in order to visually enhance the differences observed. They are major in 5C1 while, according to the blue color dominance in 5C2, they are close to zero. The aortic arch is displayed without LCCA and LSA overlapping both in the simulated projection based on expert setting and on the angle estimated by our approach.
Discussion
As an option for patients with type B aortic dissection, TEVAR has been proved a sound technique to replace open surgery [19, 20]. However, its efficacy depends on whether the proximal entry tear is completely covered or not. Therefore appropriate stent selection and deployment techniques are required. In addition, undesirable results such as occlusion or endoleak could result from a suboptimal stent deployment caused by an unreliable viewing angle [21, 22]. During the stent placement, the optimal viewing angle is subjectively selected by adjusting the rotation angle of the Xray gantry. This “trialanderror” approach increases the dose of contrast agent and radiation exposure to the doctors and patients. Besides, due to the patientspecific aortic patterns, there is no guarantee that the angle chosen by the less experienced doctor could visualize the aorta optimally during the stent deployment [23, 24]. Therefore, computerassisted selection of the optimal viewing angle is of great significance to support junior doctors during TEVAR, especially with the increasing complexity of interventional treatment for aortic dissection.
In some cases, the identification of optimal viewing angles based on 2D angiographic projections is extremely challenging. Since the reconstruction of the whole aorta in angiographic acquisitions is highly timeconsuming and needs a high quality angiographic image system [7], it appears difficult to apply this approach in routine clinical practice. In order to propose an effective and practical solution, we only need to reconstruct the thoracic aorta using the CTA data acquired before the intervention. Once the 3D reconstruction has been completed, the information can be used to plan the interventionrelated image acquisition, which will benefit the patients who need repeated radiation exposure.
The projection overlapping rate and projection foreshortening rate are two keys factors in the optimum angle determination. The former represents the overlap of the branches on the aortic arch, which is clinically used to avoid any stent covering the branches and to guide the stent placement in the endovascular procedure. As a functional estimation, the latter indicates whether the aortic arch displays the anatomical information clearly at this viewing angle. The proposed method utilizes both the PFR and POR. Instead of considering a joint minimization with the PFR and POR terms, we have shown how we can first deal with branch overlaps and then look among the solutions the optimal view minimizing any possible foreshortening. No significant difference is observed between our results and the Carm angle setting given by the experts during TEVAR.
The determination of Carm angle is necessary for patients who require endovascular surgery. The clinically used “trialanderror” approach, although efficient, increases the contrast agent dose and radiation exposure to the doctors and patients. The guiding principle of radiation safety (i.e. ALARA for “as low as reasonably achievable”) recommends reducing in any way the radiation even if the gain is small. Compared with the “trialanderror” approach, our algorithm only makes use the CTA data acquired before TEVAR to get the optimal viewing angle. The results are very close to the angle chosen by the experts and no statistically significant difference is observed. Therefore, the Carm angle can directly use the result obtained by our algorithm before the intervention. After the contrast injection, if the angle is suitable, the stent can be directly implanted. Even if the angle needs to be finetuned, the slight error will reduce the radiation time compared to the previous angle adjustment from zero. Our future work will focus on the overall benefit of our approach in terms of radiation amount and contrast usage by either experts or junior physicians.
Conclusion
In this paper, an adaptive optimization algorithm based on CTA data acquired before TEVAR is proposed to determine the clinically best Carm angle to use during the intervention. The optimal viewing angle estimated using this method does not show a significant difference with expert settings. Hence, this automatic method has the potential to assist junior doctors while providing shorter procedure time, less radiation exposure and contrast injection.
Availability of data and materials
The datasets used or analyzed during the current study is available from the corresponding author on reasonable request.
References
Qin YL, Deng G, Li TX, Wang W, Teng GJ. Treatment of acute typeb aortic dissection thoracic endovascular aortic repair or medical management alone? Jacc Cardiovasc Inte. 2012;6:185–91.
Fattori R, Cao P, Rango PD, Czerny M, Evangelista A, Nienaber C, Rousseau H, Schepens M. Interdisciplinary expert consensus document on management of type b aortic dissection. J Am Coll Cardiol. 2013;61:1661–78.
Afifi RO, Sandhu HK, Leake SS, Boutrous ML, Kumar V, Azizzadeh A, CharltonOuw KM, Saqib NU, Nguyen TC, Miller CC, Safi HJ, Estrera AL. Outcomes of patients with acute type b (DeBakey III) aortic dissection: a 13Year, singlecenter experience. Circulation. 2015;132:748–54.
Nienaber CA, Clough RE, Sakalihasan N, Suzuki T, Gibbs R, Mussa F, Jenkins MP, Thompson MM, Evangelista A, Yeh JSM, Cheshire N, Rosendahl U, Pepper J. Aortic dissection. Nat Rev Dis Primers. 2016;6:16053.
Fattori R, Mineo G, Eusanio MD. Acute type b aortic dissection: current management strategies. Curr Opin in Cardiol. 2011;6:488–93.
Andersen ND, Keenan JE, Ganapathi AM, Gaca JG, Hughes GC. Current management and outcome of chronic type b aortic dissection: results with open and endovascular repair since the advent of thoracic endografting. Ann Cardiothorac Sur. 2014;3:264–74.
Sato Y, Araki T, Hanayama M, Naito H, Tamura S. A viewpoint determination system for stenosis diagnosis and quantification in coronary angiographic image acquisition. IEEE T Med Imaging. 1998;17:121–37.
Sakamoto S, Kiura Y, Shibukawa M, Ohba S, Arita K, Kurisu K. Subtracted 3d ct angiography for evaluation of internal carotid artery aneurysms: comparison with conventional digital subtraction angiography. Am J Neuroradiol. 2006;27:1332–7.
Canaud L, Ozdemir BA, Patterson BO. Retrograde aortic dissection after thoracic endovascular aortic repair. J Vasc Surg. 2014;59:3889–95.
Dumay A, Reiber J, Gerbrands JJ. Determination of optimal angiographic viewing angles: basic principles and evaluation study. IEEE T Med Imaging. 1994;13:13–24.
Christiaens J, Walle R V D, Lemahieu I. A simple determination system for optimal angiographic viewing angles and QCA parameters. International conference on image processing. 2001. p. 327–330.
Tu S, Hao P, Koning G, Wei X, Reiber JHC. In vivo assessment of optimal viewing angles from xray coronary angiography. EuroIntervention. 2011;7:112–20.
Reiber J H C, Marquering H A, Kitslaar P H, Jukema W J, Koning G, Vossepoel A M, Nieber M, Bax J J. Automated determination of optimal angiographic viewing angles for coronary artery bifurcations from cta data. Conference on medical imaging. 2008.
Wang X, Yang J, Chen Y, Ai D, Hu Y, Wang Y. Optimal viewing angle determination for multiple vessel segments in coronary angiographic image. IEEE T Nucl Sci. 2014;61:1290–303.
Wang X, Yang J, Yang T, Wang Y. Global optimization of optimal angiographic viewing angles for coronary arteries with multiple segments. Conference of the IEEE engineering in medicine and biology society. 2013. p. 2640–2643.
Koka V, TheriaultLauzier P, Xiong TY, BenShoshan J, Petr R, Laboš M, Buithieu J, Mousavi N, Pilgrim T, Praz F. Optimal fluoroscopic projections of coronary ostia and bifurcations defined by computed tomography coronary angiography. JACCCardiovasc Inte. 2020;13:2560–70.
Chen SJ, Carroll JD. 3D reconstruction of coronary arterial tree to optimize angiographic visualization. IEEE Trans Med Imaging. 2000;19:318–36.
Zhao X R, Yang G Y, Chen, Y, Lv T L, Sun W Y, Shu H, Haigron P. Segmentation of aorta dissection ct images using convolution neural networks. In 33rd international congress and exhibition of computer assisted radiology and surgery. 2019.
Stanley GA, Murphy EH, Knowles M, Ilves M, Jessen ME, Dimaio M, Modrall JG, Arko FR. Volumetric analysis of type b aortic dissections treated with thoracic endovascular aortic repair. J Vasc Surg. 2011;54:985–92.
Nienaber CA, Kische S, Rousseau H. Endovascular repair of type b aortic dissection: longterm results of the randomized investigation of stent grafts in aortic dissection trial. J Vasc Surg. 2014;59:554.
Totaro M, Miraldi F, Fanelli F, Mazzesi G. Emergency surgery for retrograde extension of type b dissection after endovascular stent graft repair. Eur J CardioThorac. 2001;20:1057–8.
Neuhauser B, Greiner A, Jaschke W, Andreas C, Gustav F. Serious complications following endovascular thoracic aortic stentgraft repair for type b dissection. Eur J CardioThorac. 2008;33:58–63.
Tipper G, UKingIm J, Price S J, Trivedi R A, Cross J J, Higgins N J, Farmer R, Wat J, Kirollos R, Kirkpatrick P J, Antoun N M. Detection and evaluation of intracranial aneurysms with 16row multislice ct angiography. Clin Radiol. 2005; 60: 565–572.
Hyde DE, Fox AJ, Gulka I, Kalapos P, Holdsworth DW. Internal carotid artery stenosis measurement: comparison of 3d computed rotational angiography and conventional digital subtraction angiography. Stroke. 2004;35:2776–81.
Acknowledgements
Not applicable.
Funding
This work was supported by the National Science and Technology Major Project of the Ministry of Science and Technology of China (Grant No. 2018ZX10201002003, funder: GY), and the National Natural Science Foundation of china (Grant No. 61871117, funder: YC). National Science and Technology Major Project of the Ministry of Science and Technology of China had a role in the design of the study, collection, analysis, and interpretation of data. National Natural Science Foundation of china contributed in acquisition and writing the manuscript.
Author information
Authors and Affiliations
Contributions
WS proposed the research methods, collect the dataset, and wrote the manuscript. GY solved some complicated problems regarding the optimization process. YC and HS participated in designing and revising this paper. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Ethics approval and consent to participate
Our study was approved by the Ethics Committee of the Chinese PLA General Hospital. All procedures in our study adhered to the principles of the Declaration of Helsinki. Informed consent was obtained from all individual participants included in the study.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Sun, W., Yang, G., Chen, Y. et al. An adaptive optimal viewing angle determination algorithm for TEVAR operation. BMC Med Imaging 21, 141 (2021). https://doi.org/10.1186/s12880021006763
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12880021006763