DEVELOPMENT AND IMPLEMENTATION OF NOVEL STRATEGIES TO EXPLOIT 3D ULTRASOUND IMAGING IN CARDIOVASCULAR COMPUTATIONAL BIOMECHANICS

Introduction In the past two decades, major advances have been made in cardiovascular diseases assessment and treatment owing to the advent of sophisticated and more accurate imaging techniques, allowing for better understanding the complexity of 3D anatomical cardiovascular structures1. Volumetric acquisition enables the visualization of cardiac districts from virtually any perspective, better appreciating patient-specific anatomical complexity, as well as an accurate quantitative functional evaluation of chamber volumes and mass avoiding geometric assumptions2. Additionally, this scenario also allowed the evolution from generic to patient-specific 3D cardiac models that, based on in vivo imaging, faithfully represent the anatomy and different cardiac features of a given alive subject, being pivotal either in diagnosis and in planning guidance3. Precise morphological and functional knowledge about either the heart valves’ apparatus and the surrounding structures is crucial when dealing with diagnosis as well as preprocedural planning4. To date, computed tomography (CT) and real-time 3D echocardiography (rt3DE) are typically exploited in this scenario since they allow for encoding comprehensive structural and dynamic information even in the fourth dimension (i.e., time)5,6. However, owing to its cost-effectiveness and very low invasiveness, 3D echocardiography has become the method of choice in most situations for performing the evaluation of cardiac function, developing geometrical models which can provide quantitative anatomical assessment7. Complementing this scenario, computational models have been introduced as numerical engineering tools aiming at adding qualitative and quantitative information on the biomechanical behavior in terms of stress-strain response and other multifactorial parameters8. In particular, over the two last decades, their applications have been ranging from elucidating the heart biomechanics underlying different patho-physiological conditions9 to predicting the effects of either surgical or percutaneous procedures, even comparing several implantation techniques and devices10. At the early stage, most of the studies focused on FE modeling in cardiac environment were based on paradigmatic models11–15, being mainly exploited to explore and investigate biomechanical alterations following a specific pathological scenario or again to better understand whether a surgical treatment is better or worse than another one. Differently, nowadays the current generation of computational models heavily exploits the detailed anatomical information yielded by medical imaging to provide patient-specific analyses, paving the way toward the development of virtual surgical-planning tools16–19. In this direction, cardiac magnetic resonance (CMR) and CT/micro-CT are the mostly accomplished imaging modality, since they can provide well-defined images thanks to their spatial and temporal resolutions20–25. Nonetheless, they cannot be applied routinely in clinical practice, as it can be differently done with rt3DE, progressively became the modality of choice26 since it has no harmful effects on the patient and no radiopaque contrast agent is needed. Despite these advantages, 3D volumetric ultrasound imaging shows intrinsic limitations beyond its limited resolution: i) the deficiency of morphological detail owing to either not so easy achievable detection (e.g., tricuspid valve) or not proper acoustic window, ii) the challenge of tailoring computational models to the patient-specific scenario mimicking the morphology as well as the functionality of the investigated cardiac district (e.g., tethering effect exerted by chordal apparatus in mitral valve insufficiency associated to left ventricular dilation), and iii) the needing to systematically analyse devices performances when dealing with real-life cases where ultrasound imaging is the only performable technique but lacking of standardized acquisition protocol. Main findings In the just described scenario, the main aim of this work was focused on the implementation, development and testing of numerical strategies in order to overcome issues when dealing with 3D ultrasound imaging exploitation towards predictive patient-specific modelling approaches focused on both morphological and biomechanical analyses. Specifically, the first specific objective was the development of a novel approach integrating in vitro imaging and finite element (FE) modeling to evaluate tricuspid valve (TV) biomechanics, facing with the lack of information on anatomical features owing to the clinically evident demanding detection of this anatomical district through in vivo imaging. • An innovative and semi-automated framework was implemented to generate 3D model of TV, to quantitively describe its 3D morphology and to assess its biomechanical behaviour. At this aim, an image-based in vitro experimental approach was integrated with numerical models based on FE strategy. Experimental measurements directly performed on the benchmark (mock circulation loop) were compared with geometrical features computed on the 3D reconstructed model, pinpointing a global good consistency. Furthermore, obtained realistic reconstructions were used as the input of the FE models, even accounting for proper description of TV leaflets’ anisotropic mechanical response. As done experimentally, simulations reproduced both “incompetent” (FTR) and “competent-induced” (PMA), proving the efficiency of such a treatment and suggesting translational potential to the clinic. The second specific aim was the implementation of a computational framework able to reproduce a functionally equivalent model of the mitral valve (MV) sub-valvular apparatus through chordae tendineae topology optimization, aiming at chordae rest length arrangement to be able to include their pre-stress state associated to specific ventricular conformation. • We sought to establish a framework to build geometrically tractable, functionally equivalent models of the MV chordae tendineae, addressing one of the main topics of the computational scientific literature towards the development of faithful patient-specific models from in vivo imaging. Exploiting the mass spring model (MSM) approach, an iterative tool was proposed aiming to the topology optimization of a paradigmatic chordal apparatus of MVs affected by functional regurgitation, in order to be able to equivalently account for tethering effect exerted by the chordae themselves. The results have shown that the algorithm actually lowered the error between the simulated valve and ground truth data, although the intensity of this improvement is strongly valve-dependent.Finally, the last specific aim was the creation of a numerical strategy able to allow for patient-specific geometrical reconstruction both pre- and post- LVAD implantation, in a specific high-risk clinical scenario being rt3DE the only available imaging technique to be used but without any acquisition protocol. • We proposed a numerical approach which allowed for a systematic and selective analysis of the mechanism associated to intraventricular thrombus formation and thrombogenic complications in a LVAD-treated dilated left ventricle (LV). Ad-hoc geometry reconstruction workflow was implemented to overcome limitations associated to imaging acquisition in this specific scenario, thus being able to generate computational model of the LV assisted with LVAD. In details, results suggested that blood stasis is influenced either by LVAD flow rate and, to a greater extent, by LV residual contractility, being the positioning of the inflow cannula insertion mandatory to be considered when dealing with LVAD thrombogenic potential assessment