DOI: 10.1161/circ.148.suppl_1.15082 ISSN: 0009-7322

Abstract 15082: Novel Imaging/Hemodynamic Based Right Ventricular Phenotypes in Patients With Pulmonary Hypertension

Alexandra M Janowski, Scott H Visovatti, raymond Benza, Rebecca Vanderpool
  • Physiology (medical)
  • Cardiology and Cardiovascular Medicine

Introduction: Right ventricular function and ventricular dynamics associate with outcomes in pulmonary hypertension (PH). Measures of RV strain describe ventricular mechanics but are not included in RV function assessments. This study aimed to develop novel RV imaging-hemodynamic phenotypes using unsupervised machine learning.

Methods: Participants with cardiac magnetic resonance (CMR) and right heart catheterization data were identified from The Ohio State University CMR PH registry (n=81). RV areas were segmented from CMR 4-chamber images and used to determine RV free wall (RVFW) and septal strain. K-medoids clustering was used on imaging features, RV hemodynamics, and RV function (Figure A). Correlation analysis was used to reduce collinearity. Significance was measured with Kruskal Wallis and chi-squared tests. A p-value < 0.05 was considered significant.

Results: Participants were assigned to three clusters and top contributors to cluster assignment were RVEF, RVFW strain, and Ea. RV phenotypes progress from mild/no RV dysfunction (Cluster 1) to severe RV dysfunction (Cluster 3) (Figure A). Cluster 1 had normal RV function with high RVEF and PA compliance, preserved LVEF, low RVSP, and low Eed. Compared to cluster 1, cluster 2 demonstrates more RV dysfunction with reduced RVEF, stroke volume, and PA compliance with an increase in RVSP but no change in Eed. In addition to more RV dysfunction (decreased RVEF and stroke volume), cluster 3 had increased RA pressure and Eed. In terms of RV strain, (Figure B), cluster 2 had less negative RVFW and septal strain compared to cluster 1. In cluster 3, RV septal strain is similar to cluster 2 but with decreased RVFW strain.

Conclusions: Imaging and hemodynamic RV variables contribute to identification of novel RV-centric phenotypes and should be included in phenotyping efforts to better describe altered ventricular dynamics contributing to RV dysfunction in PH.

More from our Archive