The main part of the flow loop setup is represented by a 3D-printed flexible thoracic aorta including the ascending aorta, the aortic arch and the descending aorta. The model creation workflow followed a previously published work [17]. Briefly, an anonymized contrast-enhanced CT dataset of a patient who had an indication for surgical aortic valve replacement with a 25 mm prosthesis, was segmented to extract the ascending aorta, aortic arch, aortic root and supra-aortic vessels. Different datasets were measured retrospectively to select a patient sized for a 25 mm aortic valve. Exclusion criteria were any of the following in the region of interest: poor image quality (i.e. device-related artefacts), pathologic diameter change, calcifications outside the aortic root and non-standard configuration of supra-aortic vessels. After segmentation of the blood volume, the digital model was hollowed by adding a constant wall thickness of 2.5 mm external to the blood volume [17]. All vessel ends were modified in a circular uniform diameter for easy attachment to standardized connectors (Fig. 1A). The proximal end of the left ventricular outflow tract was prolonged to allow for adequate sealing, as well as placement of the heart valve prostheses according to manufacturer’s specifications. Afterwards, the digital model was transferred into the slicing software Modeling Studio (Keyence Corp., Osaka, JP), subsequently uploaded onto a 3D-printer (Agilista 3200W, Keyence Corp.) and printed using a flexible, printing material (AR-G1L, Shore 35A, elongation at break: 160%, Keyence Corp.). After the printing process, the aortic phantom was taken from the build plate and soaked in boiling water to remove the water-soluble support material. Subsequently, the model was placed in a heating cabinet to dry for 24 h at 50 °C.

3D-Printed Arch and Valve Implantation; A: Digital model of the anatomical aortic arch with straightened connectors for improved implementation in the flow loop. B: Magna Ease biological valve sutured onto the customized valve holder. C: Valve placed inside the aortic arch with opening for visualization purposes

To perform standardized comparative tests of different heart valve prostheses, a uniform prosthesis size of 25 mm (manufacturer’s specification) was selected for all valves tested in this study. Included are five different valves for surgical implantation, with two mechanical prosthetic valves (Masters Series 25, Abbott Laboratories, Chicago, USA; On-Xane-25, CryoLife Inc., Kennesaw, USA) and three different bioprosthetic heart valves (Epic 25 mm, Abbott Laboratories; Magna Ease 25 mm, Edwards Lifesciences Inc., Irvine, USA; Perimount 25 mm, Edwards Lifesciences Inc.). Individual valve mounts were designed to follow the individual curvature of the valve’s suture rings (Fig. 1B). Subsequently, valves were fixed to the mount using surgical sutures (Prolene 5–0, Ethicon Inc., Raritan, USA) and tested for paravalvular leakages. Each mount has a defined height, to allow for supra or intra-annular placement of the valves, according to manufacturer’s recommendations (Fig. 1C). The orientation of the mechanical valve leaflets was adjusted to match manufacturer’s recommendations. Bioprosthetic valves were stored in their original container with storage solution up until testing.

To allow for testing of the valves in an MRI setting, an entire MRI-compatible mock circulation setup was designed and constructed (Fig. 2). The setup was divided into two parts, the external drive unit and the internal fluid circulation unit. The external drive unit consisted of a dedicated computer, linear motor (PS01- 48 × 240 HP, NTI AG, Spreitenbach, CH) with corresponding driver (Series C1100, NTI AG). The linear motor was connected to a piston, which in turn is connected air-tight via a pneumatic hose to the fluid circulation unit. The connecting point also represents the heart of the mock circulation with a self-developed pump chamber, representing the left ventricle. To transfer the pneumatic force created by the piston to the test fluid, a rubber roll membrane with a defined volume of 80 ml was placed between the pneumatic and fluid chambers. The fluid chamber has a total volume of 100 ml resulting in a theoretical peak ejection fraction of 80%. An ejection fraction above physiological levels was chosen to adjust for the rigid nature of the artificial ventricle The chamber was connected to the valve mount via a straight rigid tube to allow for any flow disturbances to subside before passing through the valve prostheses. The 3D-printed aortic arch was then fixed to the valve mount which was placed in a plastic container. The container has five openings, for the proximal fluid entrance, the distal descending aorta and the supra-aortic vessels. After implantation of the valves in the aortic arch, the model was embedded in a hydrogel of 1% agar (Agarose, Sigma-Aldrich Corp., St. Louis, USA) to simulate the surrounding tissue and thereby reduce movement artefacts during MRI acquisition. Distal to the descending aorta and the supra-aortic vessels, a combination of compliance and resistance elements were placed to allow the approximation of the Windkessel-effect and peripheral vascular resistance. The compliance elements consist of an airtight cylinder filled partially with water and air, with a pneumatic valve at the top to adjust the height of the water column. The resistance element is realized through a ball valve that is placed distally to the compliance element. Therefore, realistic pressure conditions of 120/80 mmHg and a cardiac output of 4.6 l/min were achieved. Pressure was measured at the left ventricle, compliance chamber and descending aorta prior to MRI experiments. For all experiments, heart rate was set at 55 bpm, while systolic and diastolic pressure were adjusted to reach 120/80 mmHg. An ECG trigger signal was created and connected to the MRI according to manufacturer’s specifications. The trigger signal allowed the prospective synchronization of the ventricle movement with the acquisition time window. To simulate the viscous behavior of blood, a blood mimicking fluid (calculated viscosity 4.6cP) consisting of 40% glycerin (Rotipuran® ≥ 99.5%, Carl Roth GmbH, Karlsruhe, GER) and 60% distilled water was used [18].

Schematic of the mock circulatory loop with an external pulsatile pump and MRI trigger. The inner loop consists of a newly developed ventricle, an airtight cylinder functioning as a compliance chamber (C), as well as several ball valves being utilized as resistance elements (R) to create physiological conditions

Acquisition of the 4D-Flow MRI imaging was performed on a 1.5 T scanner (MAGNETOM Aera, Siemens Healthineers AG, Erlangen, GER) with an 18-channel body coil (Biomatrix Body 18, Siemens Healthineers AG) placed on top of the agar filled plastic box. The acquisition protocol consisted of a non-contrast-enhanced MR-angiography and the 4D-flow sequence. For 4D-flow an isotropic dataset with 25 phases and a slice thickness of 1.0 mm (TE 2.300, TR 38.800, FA 7°, matrix size: 298 × 298 px) was acquired. Velocity encoding was set at 150 cm/s for all measurements [19]. Evaluation and visualization of 4D-Flow MRI results was conducted using a dedicated radiological analysis software (cvi42, CCI Inc., Calgary, CA) [17]. Within the software, the blood volume was separated from surrounding motion artefacts. Four measurement planes were placed perpendicular to the vessel’s centerline, specifically proximal to the valve as a reference plane, 10 mm distal to the top of the valve, at the center of the ascending curvature and at the distal end of the aortic arch (Fig. 7). At each plane, velocity, tangential WSS and pressure drop with respect to the reference plane were measured. Calculation of WSS followed the publication by Stalder et al.[20]. It describes an interpolation of local velocity vectors along the contour of the underlying measuring plane. The effective orifice area (EOA) was calculated using the continuity equation (Eq. 1) with the velocity time integral in the left ventricular outflow tract (LVOT) and aortic valve (AV) derived from the underlying MRI dataset.

$$EOA= \frac_^* \frac*_}_}$$

(1)

Equation 1: Continuity equation to determine the EOA; d = diameter; VTI = velocity time integral.

Sonographic imaging was performed using a dedicated sonography device (Resona 9, Mindray Medical Int. Ltd., Shenzhen, CN) and the v-flow protocol, developed for carotid artery imaging. For image acquisition a linear array transducer (L14-3WU, Mindray Medical Int. Ltd.) was placed on to the agar block in correspondence to the above-mentioned planes, placing the center of the transducer on the according plane. The acquisition window was increased to the biggest possible size (20 × 30 mm) while all other parameters were set to the most precise setting available (acquisition time: 2 s; acquisition quality: 7). Since the acquisition window was developed for application at the carotid bifurcation, measurements had to be split into two parts at the inner and outer curvature of the aorta to cover the entire cross-section, due to the smaller ROI of the acquisition window. Flow velocity, total WSS at five spots along the aortic wall as well as the oscillatory shear index (OSI) were calculated from the measurements. The OSI was calculated as an expression for the magnitude and change in direction of local WSS described by the following formula:

$$OSI= \frac*(1.0- \frac)$$

(2)

where AWSSV = magnitude of the time-averaged WSS vector, and AWSS = time-averaged WSS magnitude [21].