Introduction

The Aortic valve (AV) is the cardiac valve that has to cope with the biggest amount of stress, and therefore is the one more frequently developing pathologies1. The AV conditions typically increase with ageing, resultant of a calcification of the leaflets (rheumatic AV) or annulus resulting in a narrowed AV area (stenotic valve). The AV can also be congenitally bicuspid. All these conditions generate an extra load to the heart. Ultimately, the AV can be repaired/replaced with clinical intervention, with the associated inherent risks. The main clinical-decision parameter to access the AV condition is the trans-valvular pressure drop (∆P), which is assessed non-invasively by simplified Bernoulli equation (SB=4*v2) based on the peak velocity across Av or invasively as the peak-to-peak (PtP) ΔP between LV and ascending aorta. With the continuous development of 4D-flow MRI, it becomes possible to acquire the flow across AV in 3D and over time with good resolution. Therefore, it is possible to account for a full velocity profile across the vena-contracta in the estimation ΔP via the simplified advective work-energy relative pressure (SAW = 1/Q (A), where Q is flow and A is the advective energy rate)2. Moreover, typical cardiovascular MRI scanning include 4D flow sequences can take for up to 45 minutes. For the elderly population, typical in AS is prevalent this can preclude investigation. Therefore, in vitro phantom models which can closely replicate valvular and aortic physiology and function can allow for intervention planning, valve assessment and potentially evaluate novel imaging techniques without the need for in vivo subject participation. The authors describe a compliant, functional valve phantom within a compliant aortic phantom under pulsatile flow to aiming to provide a realistic model of several AV conditions and compare invasive catheterisation against 4D-flow-MRI velocity ΔP.

Methods

The fabrication of compliant valves mimicking healthy AV and three AV conditions (rheumatic, stenotic and bicuspid) was based on the segmentation of computer tomography image of a patient-specific AV. Then the valve mould was meshed and 3D-printed. De-gassed silicone 0030 was injected into the mould and cured for 2 days. Each valve was inserted in the the aortic root of a compliant aorta phantom immersed in Agar gel and connected to a hydraulic circuit filled with a blood-mimicking fluid and activated by an MRI-compatible flow pump (Figure 1)3. For mirroring cardiac output from rest to stress, each valve was imagined under three pulsatile flow rates reaching 150ml/s, 200ml/s and 250ml/s (respectively except the later in the stenotic AV to avoid overcoming the maximum rated pressure of the phantom). In each flow condition, the PtP ∆P across the valve was measured by fluid-filled catheters inserted before and after the valve. Phase-contrast MRI 4D-Flow was acquired for each valve under each flow condition. The aortic phantom was segmented to define the velocity filled of view. The velocity profiles were obtained at the vena-contracta, were the basis of SB and SAW computation over the full pulsatile cycle. Linear regression and Bland-Altman plots between each methodological non-invasive ΔP estimations and the invasive ΔP measurements were investigated.

Results

In Figure 3 it is summarized the maximal pressure drop across the different implanted valves, under each flow conditions, for catheterization measurements and MRI-derived estimations (SB and SAW). The ΔP variation over time at the vena-contracta for the highest flow rate conditions is presented in Figure 4. SAW ΔP shows better agreement with invasive pressure recordings than SB ΔP which is the basis of clinical routine (correlation R2 of 0.808 versus 0.751 respectively, with a linear regression coefficient β of 0.81 versus 1.51) (Figure 5A). Also, the Bland-Altman analysis indicates that the comparison between SAW ΔP and invasive measurements have better agreement with lower limits of agreement (Figure 5B&C).

Discussion

The generation of the 3D-printed compliant valves and the respective insertion in the compliant aorta phantom allowed a comprehensive 4D-flow MRI study of the ΔP estimation based on the flow profiles very similar to in-vivo. The results show a very good agreement between the ΔP estimated via SAW in the vena-contracta and the PtP ΔP measured invasively. Simultaneously, it was further confirmed that the peak ΔP via SB is an overestimation of the real ΔP and has higher variability. These differences on the Velocity-based ΔP (SB and SAW) indicate that accounting for full velocity profile provides a more widespread, and thus accurate, assessment of ΔP than accounting simply for peak velocity used clinically. The potential impact of this study in clinical practise is valuable in 2 directions. Firstly, it provides extra evidence that the non-invasive estimations of AV ΔP should be improved. Secondly, it sets a framework to the evaluation of more complex and/or implantable personalized AVs in a fluid circuit outside of the human body and therefore risk-free, before any actual intervention, and potentially improving the respective outcome.

Conclusion

The present study provides in-vitro evidence that non-invasive ΔP across aortic valves (healthy and diseased) is better predicted by considering the full 3D flow profile at the vena-contracta. the results further prove that peak velocity across the valve overestimates aortic valve ΔP. Simultaneously, it was set a framework for generating and evaluating personalized compliant valves which are MRI-compatible.

Acknowledgements

PIC project, European Union’s Horizon 2020 Marie Skłodowska-Curie ITN Project under grantagreement No 764738. PL holds a Wellcome Trust Senior Research Fellowship (g.a. 209450/Z/17/Z).