MechE PhD Prospectus Defense: Hiba Kobeissi

TITLE: Data-Driven Approaches to Understanding the Mechanical Behavior of Soft Materials with an Emphasis on Cardiac Tissue

ABSTRACT: Modeling biological soft tissue is very challenging in part due to their material heterogeneity dictated by complex microstructural patterns, which play a major role in defining the mechanical behavior of these tissues. Experimentally characterizing these patterns is often a very intricate and laborious task, and computationally modeling their behavior in a faithful manner can also be extremely challenging and costly. In this context, my research interests and work lie in the realm of exploring unconventional data-driven methods to understand mechanical aspects of tissue behavior and hopefully, contribute to ongoing endeavors to better characterize tissue spatial heterogeneity. Recently, there has been growing interest in applying machine learning (ML)--based methods to predict the mechanical behavior of heterogeneous materials and closely approximate computationally expensive heterogeneous material simulations. However, when it comes to application on soft tissue, the number of useful examples available to train a model is often limited. This limitation motivated the work presented in Chapter 1, where we investigated the efficacy of both ML--based generative models and procedural methods as tools for augmenting limited input pattern datasets and added a dataset of heterogeneous material domains to the open--source Mechanical MNIST collection. Moving forward, we are currently working on studying live soft tissue, specifically human induced pluripotent stem cell derived cardiac microbundles. These lab-grown cardiac tissues have many applications in different research areas, including disease modeling, drug discovery, and regenerative tissue engineering. However, non-destructively characterizing their beating behavior and in a high throughput manner is a major challenge in the field. In this regard, we develop a computational framework for automated quantification of morphology-based mechanical metrics from movies of cardiac microtissue. Our framework relies solely on the natural contrast of the microtissue textures and computes full-field displacements and subdomain-averaged strains among other results. We elaborate on this work in Chapter 2 and provide a number of preliminary results. Once finalized, we aim to release this software as an open--source computational tool. Looking forward, we anticipate expanding the capabilities of this software to analyze other components of these microtissue movies, specifically, the pillars to which the tissues are attached. We also aim to adapt our current framework which accepts brightfield movies to include calcium imaging and derive meaningful relevant metrics. Other possible future directions include (1) using ML--based data analysis techniques to identify patterns in these datasets, and (2) using these results to directly inform computational models.

COMMITTEE: ADVISOR/CHAIR Professor Emma Lejeune, ME; Professor Paul Barbone, ME/MSE; Professor Katherine Yanhang Zhang, ME/BME/MSE