Cardiovascular dynamics in patients with pulmonary hypertension

Mentors:
Lead: Mette Olufsen (Mathematics, NC State), Collaborators: Brian Carlson (Physiology, University of
Michigan)

Outline:
Cardiovascular disease remains the leading cause of death in adults above 65. Heart failure (HF) is one of the most common causes of death, yet this disease includes many subphenotypes, several of which are not uniquely defined. As a result, diagnosing HF is challenging, and procedures aimed at identifying different disease phenotypes often require invasive measurements that carry significant risk. We will use cardiovascular measurements obtained from three different procedures: right heart catheterization, transthoracic echocardiography, and cardiac magnetic resonance imaging extracted from electronic health records from over 300 patients, made available from our collaborator at the University of Michigan. We will evaluate each patient’s cardiovascular state using physics-based and data-driven models to develop better methods for HF phenotyping. Results from this study will extend analytical, statistical, and physics-based models available to analyze quantitative data [1,2,3].

Objectives and Outcomes:
We will employ an existing physics-based cardiovascular model [5], along with sensitivity analysis, subset selection, and optimization, to determine and estimate identifiable parameters [1,3,4]. Unsupervised machine learning [3] on these identifiable parameters will be used to identify different HF phenotypes. We will also use machine learning on our data alone to evaluate differences between data- and model-driven phenotyping approaches. We will test the validity of distinguishing parameters extracted from predictive models in a moderated- mediation statistical approach through modeling exercises and sensitivity analyses.

References:
[1] A. Colunga, M.J. Colebank, REU Program, and M.S. Olufsen. Parameter inference in a computational model of haemodynamics in pulmonary hypertension. J R Soc Interface, vol. 20, p. 20220735, 2023, doi: 10.1098/rsif.2022.0735.

[2] M.J. Colebank, P.A. Oomen, C.M. Witzenburg, A. Grosberg, D.A. Beard, D. Husmeier, M.S. Olufsen, N. Chesler. Guidelines for mechanistic modeling and analysis in cardiovascular research. Am J Physiol, vol. 327, pp. H473-H503, 2014, doi: 10.1152/ajpheart.00766.2023.

[3] E. Jones, E. Randall, S. Hummel, D. Cameron, D. Beard, and B. Carlson. Phenotyping heart failure using model-based analysis and physiology-informed machine learning. J Physiol, vol. 599, pp. 4991-5013, 2021, doi: 10.1113/JP281845.

[4] N. Woodall, K. Kim, A. Colunga, J. Gennari, M. Olufsen, and B. Carlson. Deep phenotyping of cardiac function in heart transplant patients using cardiovascular system models. J Physiol vol.598, p. 3203-3222, 2020. doi: 10.1113/JP279393.

[5] A. Colunga, B. Carlson and M. Olufsen. The importance of incorporating ventricular-ventricular interaction (VVI) in the study of pulmonary hypertension. Math Biosci vol.375:109242, 2024.doi: 10.1016/j.mbs.2024.109242.