Once the positions of the valve landmarks have been initialized, automatic image-based tracking procedure of the landmarks is started as follows: first, each input image of the sequence and the contrast image have to be preprocessed to reduce the image noise by using median filter.
Second, a rigid intensity-based image registration framework between the input image as a fixed image and the contrast image as a moving image is applied, based on the Insight Segmentation and Registration Toolkit (ITK).25 The image registration framework includes four components:
* A mutual information metric by Mattes et al26 has been used to define how well the two images match each other based on similar intensity features between the input image and the predefined contrast image (Fig. 3). These matched objects are the transesophageal echocardiography probe, guide wires, the catheter, the rib spreader, and the valve delivery system.
* A linear interpolation allows intensity estimation of the moving image in a nongrid position after mapping the fixed image space onto the moving image space.
* A transformation maps the fixed image onto the moving image by resolving the translational misalignment between the two images to overlap the same objects in both images.
* A regular step gradient descent optimizer is applied with four iterations to explore the optimal values of these translational parameters in x-y coordinates that allow to update the registered image transformation in real time.
Third, the initial positions of the valve landmarks in the contrast image are updated according to the final transform parameters in current processed image. Finally, the new position of the valve landmarks is overlaid onto each image of the sequence to visualize the boundaries of the target area for the valve implantation in the 2D fluoroscopic images during the intervention.
Evaluation of the registration-based tracking procedure is a challenging task, because of the lack of contrast agent in most fluoroscopic images of the sequence. Moreover, there is no ground truth image data for the aortic root including the valve landmarks. Therefore, all fluoroscopic images have been visually assessed and qualitatively inspected to validate the quality score of distance localization errors of the target.27 The quality scores of the tracked valve landmarks are estimated as follows: (a) good is defined as an excellent alignment to a minimal misalignment between the input sequences and the registered contrast image with a difference of less than 2 mm; (b) moderate is defined when a relatively high misalignment between the two images occurs between 2.0 and 5.0 mm; and (c) insufficient is defined as a significant misalignment of more than 5.0 mm between both images.
The developed tracking method of aortic valve landmarks was tested and evaluated on different fluoroscopic image sequences of 10 patients during routine transapical TAVI. The dimensions of fluoroscopic images are 1024 × 1024 pixels, with a pixel size of approximately 0.2 mm. All fluoroscopic images sequences were acquired during routine angiographic and fluoroscopic x-ray imaging with a floor mounted C-arm system (Artis zeego, Siemens AG, Healthcare Sector, Forchheim, Germany).
Figure 4 shows a screenshot of the developed fluoroscopic guidance software for assisting the TAVI. The image-registration tracking algorithm has been implemented using C++ programming language being suitable for real-time image processing with a frame rate of 10 to 15 frames per second.
In Figure 5, two different sets of tested fluoroscopic images are presented showing successful results of valve landmark tracking, based on the intensity-based image registration. The contrast images including the manual localization of aortic valve landmarks are shown in Figures 5a, b. To visualize the matching accuracy between the contrast image (moving image) and the intensity features of each live fluoroscopic image (fixed image), checkerboard images are shown as alternating blocks from the contrast image and the input fluoroscopic image in the presence of contrast agent (Figs. 5c, d) and without contrast agent as shown in Figures 5e, f.
Significant errors of valve landmark tracking appeared in the image sequence (seq. 7) among all the tested datasets (Fig. 6). Although the contrast image has been successfully automatically detected as shown in Figure 6a, checkerboard registration visualizes the visible discrepancies between the input fluoroscopic images and the detected contrast image.
The registration-based tracking algorithm has been qualitatively evaluated for all tested images of the fluoroscopic sequences as a percentage of the total number of images per sequence as shown in Figure 7. In image sequences 6, 7, and 10, insufficiently registered landmarks were observed. The seq. 7 showed the maximum percentage of images with an insufficient score of 25% of all images per sequence, due to severe image noise and a small number of similar intensity features between the contrast image and each image of the sequence. The moderate score of landmarks tracking did not exceed 40% in seq. 5 among all the tested image sequences. However, most images of the 10 fluoroscopic sequences showed successful tracking of landmarks with a good score between 60% and 95% of all fluoroscopic images, such as sequences 2, 3, and 8.
The positioning of the AVP is the most critical step of the transapical TAVI procedure under 2D fluoroscopy guidance. Avoiding usage of additional contrast agent is a key factor to solve the current limitations of single-plane x-ray fluoroscopy for the image guidance of TAVI interventions.
We demonstrated in this study that 2D fluoroscopic image registration is potentially a feasible method for intraoperative tracking of the aortic valve landmarks to assist the position of the AVP without further contrast agent injections. Moreover, the main advantage of image-based tracking methods over optical and electromagnetic tracking systems is that image-based guidance systems do not require expensive instrumentation or sophisticated imaging technologies. Therefore, our image register tracking-based algorithm provides an appropriate solution for real-time tracking of anatomic valve landmarks in live 2D fluoroscopic images.
The qualitative evaluation of the registration performance showed that the quality of registration based on landmark algorithm tracking is good as long as the alignment errors are less than 2.0 mm as seen in most tested sequences. Three of 10 fluoroscopic image sequences showed misalignment in the range of 1% to 25% insufficient registered images per sequence. Severe image noise and the small number of similar intensity features such as transesophageal echocardiography probe and guide wires affected the image registration accuracy for seq. 7 (Fig. 6). However, the alignment of fluoroscopic images is still valid and optimized by using the capabilities of image registration framework.
We have developed a new sensorless tracking method of stenotic aortic valve landmarks for assisting the transapical TAVI under 2D fluoroscopy guidance. This method can provide a helpful tool for the surgeon by automatically defining the desired position of the AVP. In addition, the developed method is used at low levels of contrast agent injections for controlling the final placement of the prosthesis to reduce long-term negative effects, such as renal failure, in patients. Therefore, the accuracy of valve implantation will increase; surgery time can be significantly reduced while increasing the overall safety of the surgical procedure.
To minimize the user interaction and to increase the accuracy of TAVI guidance, we are currently integrating the 3D information of the aortic root model including the valve landmarks from intraoperative C-arm computed tomographic images28 with live 2D fluoroscopic images.
1. Coeytaux RR, Williams JW, Gray RN, Wang A. Percutaneous heart valve replacement for aortic stenosis: state of the evidence. Ann Intern Med.
2. Singh IM, Shishehbor MH, Christofferson RD, et al. Percutaneous treatment of aortic valve stenosis. Cleve Clin J Med.
3. Iung B, Baron G, Butchart EG, et al. A prospective survey of patients with valvular heart disease in Europe: The Euro Heart Survey on Valvular Heart Disease. Eur Heart J.
4. Bonow RO, Carabello BA, Chatterjee K, et al. 2008 Focused update incorporated into the ACC/AHA 2006 guidelines for the management of patients with valvular heart disease: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 1998 Guidelines for the Management of Patients With Valvular Heart Disease): endorsed by the Society of Cardiovascular Anesthesiologists, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. Circulation.
5. Sherif M, Tölg R, Richardt G. Percutaneous aortic valve replacement: results update, future trends and new challenges. Clin Res Cardiol Suppl.
6. Cribier A, Eltchaninoff H, Bash A, et al. Percutaneous transcatheter implantation of an aortic valve prosthesis for calcific aortic stenosis: first human case description. Circulation.
7. Grube E, Schuler G, Buellesfeld L, et al. Percutaneous aortic valve replacement for severe aortic stenosis in high-risk patients using the second- and current third-generation self-expanding CoreValve prosthesis: device success and 30-day clinical outcome. J Am Coll Cardiol.
8. Walther T, Falk V, Borger M, et al. Minimally invasive transapical beating heart aortic valve implantation—proof of concept. Eur J Cardiothorac Surg.
9. Webb JG, Pasupati S, Humphries K, et al. Percutaneous transarterial aortic valve replacement in selected high-risk patients with aortic stenosis. Circulation.
10. Walther T, Dewey T, Borger MA, et al. Transapical aortic valve implantation: step by step. Ann Thorac Surg.
11. Yan TD, Cao C, Martens-Nielsen J, et al. Transcatheter aortic valve implantation for high-risk patients with severe aortic stenosis: a systematic review. J Thorac Cardiovasc Surg.
12. Wendler O, Walther T, Schroefel H, et al. The SOURCE Registry: what is the learning curve in trans-apical aortic valve implantation? Eur J Cardiothorac Surg.
13. Al Ali AM, Altwegg L, Horlick EM, et al. Prevention and management of transcatheter balloon-expandable aortic valve malposition. Catheter Cardiovasc Interv.
14. Thomas M, Schymik G, Walther T, et al. Thirty-Day Results of the SAPIEN Aortic Bioprosthesis European Outcome (SOURCE) Registry: a European registry of transcatheter aortic valve implantation using the Edwards SAPIEN valve. Circulation.
15. Walther T, Schuler G, Borger M, et al. Transapical aortic valve implantation in 100 consecutive patients: comparison to propensity-matched conventional aortic valve replacement. Eur Heart J.
16. Van de Veire N. Imaging to guide transcatheter aortic valve implantation. J Echocardiogr.
17. Gessat M, Merk DR, Falk V, et al. A planning system for transapical aortic valve implantation. Conf Proc SPIE Med Imaging.
18. Karar ME, Gessat M, Walther T, et al. Towards a new image guidance system for assisting transapical minimally invasive aortic valve implantation. Conf Proc IEEE Eng Med Biol Soc.
19. Karar ME, Merk DR, Chalopin C, et al. Aortic valve prosthesis tracking for transapical aortic valve implantation. Int J Comput Assist Radiol Surg.
20. John M, Liao R, Zheng Y, et al. System to guide transcatheter aortic valve implantations based on interventional C-arm CT imaging. Med Image Comput Comput Assist Interv.
21. Li M, Mazilu D, Horvath K. Robotic system for transapical aortic valve replacement with MRI guidance. In: Metaxas D, Axel L, Fichtinger G, Székely G, eds. Medical Image Computing and Computer-Assisted Intervention—MICCAI 2008.
Vol. 5242. Berlin/Heidelberg: Springer; 2008:476–484.
22. Li M, Kapoor A, Mazilu D, Horvath KA. Pneumatic actuated robotic assistant system for aortic valve replacement under MRI guidance. IEEE Trans Biomed Eng.
23. Markelj P, Tomazevic D, Likar B, Pernus F. A review of 3D/2D registration methods for image-guided interventions. Med Image Anal.
24. Condurache AP, Aach T, Eck K, Bredno J. Fast detection and processing of arbitrary contrast agent injections in coronary angiography and fluoroscopy. Paper presented at: Bildverarbeitung für die Medizin 2004 (Algorithmen, Systeme, Anwendungen); March 29–30, 2004; Berlin.
25. Kitware. The Insight Segmentation and Registration Toolkit; 2010. Available at: www.itk.org
26. Mattes D, Haynor DR, Vesselle H, et al. PET-CT image registration in the chest using free-form deformations. IEEE Trans Med Imaging.
27. Kolar R, Kubecka L, Jan J. Registration and fusion of the autofluorescent and infrared retinal images. Int J Biomed Imaging.
28. Zheng Y, John M, Liao R, et al. Automatic aorta segmentation and valve landmark detection in C-arm CT: application to aortic valve implantation. Med Image Comput Comput Assist Interv.
This is a report of an innovative new technique for sensorless tracking of stenotic valve landmarks using 2D fluoroscopic imaging to assist in the accurate placement of a transapical transcatheter aortic valve. This is a timely report as transcatheter aortic valve implantation (TAVI) is frequently performed in Europe and has just been approved for inoperable patients in the U.S. Mispositioning during TAVI can have disastrous and, in rare instances, fatal consequences. The advantages of this technique include that it does not require other expensive instrumentation, such as MRI guidance, or additional usage of contrast agents. The disadvantage is that there was insufficient image tracking in 3 of 10 images, and the percentage of good tracking ranged from under 50% to 95%. Further work will be needed to define the utility of this new tracking methodology.
Keywords:©2011 by the International Society for Minimally Invasive Cardiothoracic Surgery
Aortic valve; Biomedical image processing; Image-guided interventions; Minimally invasive valvular cardiac surgery; X-ray fluoroscopy