The present study demonstrates the preclinical feasibility of entirely CMR-guided TAVI in a swine model. This was achieved using the commercially available CoreValve prosthesis without alterations and a modified, CMR-compatible delivery device. As a single imaging modality, CMR provided (a) comprehensive diagnostic evaluation of the relevant cardiac and vascular anatomy for adequate interventional planning, (b) reliable procedural guidance in real-time, (c) immediate evaluation of procedure-related complications, and (d) post-interventional validation of treatment success. The concept of rtCMR-guided TAVI was first described by Kuehne et al. in 2004. This archetype study in the preclinical, pioneering days of TAVI demonstrated the general technical feasibility and potential value of rtCMR guidance in acute animal experiments . The authors implanted entirely custom-built self-expanding, nitinol-based stent-valves into the native aortic valves of seven pigs using passive device visualization. The stent-valves were loaded into a 10 French delivery catheter with two ferromagnetic markers and successfully implanted via the carotid and iliac artery. Subsequently, McVeigh et al. described the rtCMR-guided implantation of custom-built balloon-expandable stent-valves delivered by the surgical, transapical approach in Yorkshire swine . They used conventional stentless aortic bioprostheses sutured into a commercial platinum-iridium stent and mounted onto a commercial balloon catheter. In their initial study, rtCMR-guided implantation with passive device visualization was technically successful in 6 out of 8 animals with 2 procedural deaths after valve deployment due to coronary artery obstruction and 2 additional procedural deaths during device manipulation. The same group further pursued the transapical approach focusing on chronic animal studies using their balloon-expandable valve  and, very recently, a custom-built self-expandable stent-valve for comparison . Interestingly, the authors demonstrated shorter procedure times and easier manipulation with the self-expandable device, resulting in fewer procedural complications.
These custom-built, CMR-compatible devices provide excellent visualization for rtCMR guidance. Our study benefits from the use of the commercial, Conformité Européenne certified CoreValve bioprosthesis, which is already well-tested in the clinical arena with thousands of human implantations world-wide. The original CoreValve bioprosthesis in combination with a modified, CMR-compatible delivery device was chosen on the basis of a previous comprehensive in vitro evaluation of both currently commercially available stent-valves and their delivery catheters, which was recently performed by our group .
As shown by the present study, passive device visualization using real-time TrueFISP imaging with radial k-space filling provided excellent real-time visualization of the delivery system with the mounted CoreValve prosthesis and the surrounding anatomy during (a) navigation through the vasculature towards the aortic valve, (b) aortic valve passage, (c) initial positioning, (d) device deployment, and (e) catheter withdrawal. Such steady-state sequences, when used with high excitation flip angles, provide a high blood signal even without administration of a contrast agent and thus provide good instrument-to-background contrast in the CMR image . The temporal resolution of seven reconstructed images per second, displayed without detectable image reconstruction delay on an in-room monitor, proved sufficient for direct and precise control of catheter movement. Moreover, rtCMR provided real-time monitoring of cardiac function and rapid detection of procedural complications.
In 6 of the 8 animals, a 26 mm CoreValve prosthesis was successfully placed across the native aortic annulus with implantation depths between 4 and 20 mm into the LVOT without ventricular embolization or dislocation into the ascending aorta, without coronary artery obstruction, and without structural or functional impairment of the mitral valve as confirmed by ex-vivo histology. Two implant failures occurred in our series. The first was a result of unsuccessful aortic arch passage and resulted in a controlled deployment of the stent-valve in the thoracic aorta distal to the left subclavian artery. Though considered an implant failure in this case, a controlled, safe deployment of the CoreValve prosthesis in the thoracic aorta without accidental vessel obstruction is sometimes required in clinical application when the CoreValve prosthesis dislocates into the ascending aorta during deployment and cannot be retrieved, as observed in approximately 10% of cases . The second implant failure occurred due to perforation of the left ventricular apex by the delivery device tip as a result of an avoidable operating error, which was not related to periprocedural imaging guidance. However, this complication, which is observed in ~1% of cases , was detected by rtCMR without time delay and without the need for an additional imaging modality such as echocardiography. Both cases indicate that rtCMR might improve both precision and safety of the TAVI procedure.
In addition to improved procedural guidance, CMR provided a reliable pre-interventional diagnostic evaluation before as well as an adequate post-interventional validation after valve implantation. Our study could demonstrate that high-resolution TrueFISP retro sequences enable detailed visualization of all anatomic landmarks required for TAVI and allowed precise structural evaluation of the procedural result with good accordance with autopsy findings. It should be noted here, that CMR as a 3D imaging modality can also provide a comprehensive assessment of the elliptical shaped aortic annulus in different scan plane orientations. Such a detailed, 3D assessment might potentially have an impact on future recommendations for size selection of the stent-valve that is - in current clinical practice based - on a single 2D transoesophageal echocardiographic measurement in the midoesophageal, long-axis LVOT view. In this context, Koos et al.  have recently demonstrated in a cohort of TAVI patients that this single echocardiographic measurement correlates well with both CMR and dual-source computed tomographic measurements of sagittal aortic annulus diameters, the sagittal long-axis view on CMR and computed tomography basically having the same orientation as the midoesophageal long-axis view on echocardiography. In contrast, annulus diameters by transoesophageal echocardiography were significantly smaller than coronal aortic annulus diameters by CMR and by computed tomography. Regarding the TAVI strategy, the authors found a perfect agreement between transoesophageal echocardiography and sagittal CMR or computed tomography measurements. In contrast, decision-making based on coronal CMR or computed tomography measurements would have modified the TAVI strategy in 22% and 24% of cases when compared to the echocardiography-based strategy.
We could further demonstrate that functional assessment of the implanted stent-valves can be performed non-invasively using flow-sensitive PC sequences, which are already used for flow measurements in the clinical evaluation of valvular heart disease [29, 30]. They allow detection of (para)valvular regurgitation immediately after valve implantation, which is important in clinical practice since hemodynamically relevant regurgitation after TAVI is associated with an increased in-hospital mortality  and may require immediate post-dilation of the prosthesis using a balloon or even valve-in-valve implantation. In our study, no relevant (para)valvular regurgitation was observed which may not be surprising in view of the fact that oversized prostheses were used.
Our study clearly shows the advantages of rtCMR-guided TAVI but also reveals its chief obstacle for translation into clinical application, namely the lack of suitable, CMR-compatible guidewires providing enough stiffness. When considering clinical application of rtCMR-guided TAVI, stiff guidewires are necessary. However, conventional metallic guidewires and guidewires with a nitinol core are not CMR-compatible since such long conducting structures might couple with the radiofrequency transmit energy of the body coil; this coupling could result in amplification of the local electric field and lead to excessive tissue heating, which presents a safety hazard precluding clinical application . Currently, CMR-compatible guidewires are polyetheretherketone-based [20, 32] or built from fiberglass  or other non-metallic, reinforced components and are passively visualized by their susceptibility artifacts, but do not provide mechanical stability equivalent to the commercial products used and required for TAVI (Amplatz Superstiff, Boston Scientific, MA, USA), as also shown in our study. Hence, the development of stiff guidewires suitable for CMR-guided TAVI is warranted. The development of CMR-compatible, metal-based guidewires might be an option. The precondition, however, remains that such guidewires do not act as radiofrequency antennas and must, consequently, include measures that counteract radiofrequency heating in the CMR environment.
When moving towards clinical application of rtCMR guided TAVI, other procedural and safety aspects should be taken into consideration and need to be discussed, specifically an adequate monitoring of heart rhythm and the issue of rapid right ventricular during preparatory balloon aortic valvuloplasty as well as an adequate procedural environment: A detailed ECG-monitoring is an essential prerequisite for TAVI. Arrhythmias (importantly ventricular tachycardia/fibrillation, bradycardia, atrial fibrillation), conduction abnormalities (importantly higher-degree AV-blocks, bundle-branch blocks) and signs of cardiac ischemia (importantly ST-elevation) need to be accurately assessable. The ECG-electrodes that have been used in our experimental setting were provided by the CMR scanner's manufacturer (Siemens Healthcare Sector, Erlangen, Germany) and were used for sequence triggering purposes only. With this setup, a comprehensive ECG-monitoring during a CMR-guided intervention is not possible. However, we did not observe any relevant decrease in heart rate/bradycardia suggestive of a higher-degree AV-block after TAVI. This might potentially be explained by the fact that implantation was performed into a juvenile, non-calcified, non-stenosed aortic valve. Nevertheless, a patient setting would require a CMR-compatible monitoring-equipment which is, however, already commercially available (e.g. Schiller MAGLIFE Serenity, Schiller Ottobrunn, Germany).
At present, balloon aortic valvuloplasty is generally performed as a preparatory step prior implantation of current transcatheter aortic valves. In order to prevent dislodgement of the inflated balloon into the ascending aorta as a result of cardiac contraction, rapid right-ventricular burst pacing is performed during valvuloplasty to induce a transient, functional cardiac arrest. Likewise, rapid right-ventricular pacing is needed for the implantation of the balloon-expandable Edwards prosthesis to ensure a stable position of the stent-valve during implantation. In contrast, the self-expandable CoreValve prosthesis does not require rapid pacing since its design allows for nearly continuous trans-aortic blood flow during stepwise implantation. Moreover, Grube and colleagues have recently demonstrated that CoreValve implantation is also feasible and safe without prior valvuloplasty . Rapid right-ventricular pacing is, therefore, not required for CoreValve implantation and does not present another obstacle for procedural guidance by rtCMR.
Currently, TAVI is ideally performed in so-called hybrid operating rooms that offer a sterile environment with full angiographic, anaesthesiological and surgical equipment. Such rooms not only provide optimal conditions for the procedure but also provide a rapid interdisciplinary team approach for complication management. While peripheral vascular complications might already be managed by PTA and stent implantation under CMR-guidance, other complications might occur that cannot be addressed within the CMR scanner. Coronary artery obstruction, for example, may require percutaneous coronary intervention or even surgery. Hence, the CMR scanner should ideally be integrated in the architectural concept of such a hybrid room, which must warrant a rapid evacuation of patients from the CMR scanner in order to perform conventional X-ray guided intervention or surgery without delay. This can be ensured either by a swivelling operating table if the CMR scanner is integrated directly within the hybrid room or a rail-mounted operating table if the CMR scanner is accommodated in a separate, adjacent room. Alternatively, a patient transfer board on a non-magnetic patient transporter or even a movable CMR scanner might be considered, although these approaches are clearly more time-consuming.
As a limitation of our study, it also has to be noted that patients referred for TAVI usually show somewhat calcified and tortuous vessels. These vessels may present a more complex anatomy than seen in our animals. However, the benefits of the detailed soft-tissue contrast and 3D representation provided by CMR may be even more dramatic in these patients. Safe navigation through the tortuous vascular access routes is presumably simplified by rtCMR to guide operator adjustments and visualize device-related anatomic distortion. In view of the stenotic, calcified aortic valve in TAVI patients, potential CMR imaging artifacts arising from valvular and aortic root calcifications need to be acknowledged. Basically, calcifications appear as areas of low signal intensity or signal void on CMR imaging. Such signal voids can make edge discrimination of calcified valve leaflets difficult, especially during direct planimetry of aortic valve area in a cross-sectional view. They might also affect rtCMR-guidance of aortic valve passage during TAVI. However, in clinical practice, passage of the stenotic aortic valve is performed with a straight-tip guidewire in a probing fashion, since neither conventional X-ray angiography nor transoesophageal echocardiography allow precise, targeted steering of the wire through the small orifice. Hence, these imaging artifacts may be negligible at this stage of the procedure, and they should also not relevantly impair valve positioning and implantation since the areas of signal void are typically constraint to the areas of calcification and do not lead to further image distortion, which might impair detailed visualization of the relevant anatomical structures. Thus, the landing zone for the CoreValve prosthesis would still be accurately visible. However, these considerations remain speculative since a calcific aortic valve could not be simulated in the pig model. Interestingly, however, successful implantation of an oversized CoreValve prosthesis in a native, non-stenotic, non-calcified aortic valve without tissue damage and dislocation might potentially have clinical implications for future catheter-based treatment of aortic regurgitation which is currently not an indication for TAVI.