| dc.description.abstract | Cardiac growth and remodeling (G&R) are intricate processes that occur in response to various physiological and pathological conditions to maintain the optimal function of the heart. These processes involve complex molecular, cellular, and structural alterations that influence the size, shape, and function of the heart. Research efforts, utilizing in vitro and in vivo models, have shed light on the mechanisms of G&R at the cellular and tissue levels. Clinically, we have access to a wealth of sophisticated information, ranging from 0D (e.g. pressure, heart rate, and ECG) to 4D (e.g. dynamic MRI and CT) data about our cardiovascular system. However, a significant challenge lies in the interpretation of these data, as they provide macroscale information and how this links to microscale mechanisms remains poorly understood. In consequence, a vast amount of clinical data, particularly imaging data, are employed qualitatively rather than quantitatively. Furthermore, many microstructural discoveries have not yet been fully leveraged for the improvement of diagnostic accuracy and therapeutic strategies.
The main goal of this thesis is to develop multiscale patient-specific simulations to connect microscale insights and clinically accessible macroscale information. I propose a growth characterization workflow utilizing in vivo cardiac MRI, kinematic growth modeling, and inverse finite element method to quantify tissue level growth. Tested on in vivo data of swine models, this approach showcases the potential of non-invasive measurement of microstructural growth properties. It also highlights two potential areas of improvements: a more granular growth analysis currently hindered by the simplified kinematic growth model and a more efficient process of the inverse analysis, catering to the time-sensitive nature of clinical practice. To address this, I create a multiscale model using a microstructurally motivated growth theory . The model enables investigation of changes in different tissue components (e.g. myocytes and collagen) during the process of G&R and its subsequent reversal. In addition, I develop a data-driven surrogate model that can efficiently generate dynamic patient-specific simulations of the left ventricle with mean nodal error below 3 mm. The surrogate model achieves a speed increase of 4 orders of magnitude compared to the traditional finite element model, making it ideal for inverse studies in fast-paced clinical settings.
In summary, this thesis introduces simulation frameworks that effectively quantify G&R properties, bridge the mechanobiological insights of G&R across different scales, and facilitate the generation of personalized models. These advancements represent progress towards the realization of cardiac digital twins in the realm of clinical practice. | |
