Research Interests

My research combines computational modeling and data analysis with in vivo and in vitro experiments to understand the progression of pulmonary vascular disease.

Computational cardiovascular dynamics.

Cardiovascular disease (CVD) is the leading cause of death worldwide, accounting for approximately 17.9 million fatalities each year [1]. Though advances in non-invasive and invasive technologies have improved diagnosis of CVD, there lacks a wholistic, integrative tool for understanding the structure and function of the cardiovascular system in disease. 

Computational models of the cardiovascular system remedy this issue, providing a framework that is able to integrate clinically data to forecast in-vivo dynamics. In addition, computational models can be used to understand vascular and cardiac function in both healthy and disease states and test physiological hypotheses rapidly, enabling efficient experimental design for clinical discovery. These models utilize computational fluid dynamics, soft tissue mechanics, and reduced order modeling to understand how hemodynamic factors are altered in CVD. 

https://www.who.int/health-topics/cardiovascular-diseases (Accessed 9/11/2021)

Pulmonary hypertension.

Pulmonary hypertension (PH), defined by mean pulmonary arterial pressure > 20 mmHg at rest, is a debilitating disease that compromises both the pulmonary vasculature and the right ventricle. The disease is divided into five main subgroups, only one of which is considered curable, and is common in those with left-sided heart failure. With increasing incidence rates of the disease and a potential link between pulmonary vascular disease and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), it is imperative to study the structure and function of the pulmonary circuit.

To understand the progression of PH, I utilize in silico, in vivo, and in vitro experiments capturing the development of the disease at the cell, tissue, organ, and system level. In silico simulations of vascular and cardiac physiology can be used to efficiently deduce hypothesized mechanisms of PH progression, and lead to experimental design for in vivo animal studies and in vitro cell culture studies. 

Inverse problems and uncertainty quantification.

Mathematical models of biological and physiological processes often include numerous parameters and limited measurable data. This makes the inverse problem, i.e., inferring model parameters from measured data, extremely diffuclt, and requires innovative model analysis techniques. 

This thrust of my research utilizes sensitivity analysis and uncertainty quantification techniques to deduce:

In the context of cardiovascular models, I consider both local and global sensitivity methods to identify which biomarkers of disease have the largest impact on model predictions. Identifiability methods such as profile-likelihood or Markov chain Monte Carlo (MCMC) can further deduce which parameters are identifiable given limited data. Similar techiques can be used to estimate posterior distributions for each parameter, and subsequently used to quantify uncertainties in model forecasts. These steps are necessary in determining patient-specific cardiac parameters and elucidating variability in model simulations.

Digital Twins of Human Health

The concept of a Digital Twin is largely attributed to the Apollo 13 mission. During the infamous "Houston, we have a problem," moment, astronauts aboard Apollo 13 were faced with a malfunction in their spacecraft. The physical  Apollo 13 was over 200,000 miles away, but required NASA's Houston Mission Control to provide feedback on what to do next. Hence, NASA turned to their Apollo 13 simulators as well as continuous measurement and data updates  to trouble shoot and identify a plan to save all those on board.


In a similar way, digital twins of human health require the same tools. Here, the simulator is a computational (usually mechanistic) model of the biophysical system. The simulator is sequentially updated with new information via sensors, measurements, and observations. This pipeline goes beyond just a patient-specific model, and requires multiple steps, including model development, optimal experimental design, sensor innovation, model analyses (as mentioned above), uncertainty quantification, and data assimilation. I am interested in developing these pipelines in my future work, including (but not limited to) cardiovascular digital twins.


For more insight, consider this webpage on the Apollo 13 mission as well as the recent book A Toolbox for Digital Twins: From Model-Based to Data-Driven by Mark Asch.