Right Ventricular Shape Feature Quantification for Evaluation of Pulmonary Hypertension: Feasibility and Preliminary Associations With Clinical Outcome Submitted for Publication

Autor: Jing, Xu, Eleanor L, Desmond, Timothy C, Wong, Colin G, Neill, Marc A, Simon, John C, Brigham
Rok vydání: 2021
Předmět:
Zdroj: Journal of biomechanical engineering. 144(4)
ISSN: 1528-8951
Popis: This study aimed to demonstrate feasibility of statistical shape analysis techniques to identify distinguishing features of right ventricle (RV) shape as related to hemodynamic variables and outcome data in pulmonary hypertension (PH). Cardiovascular magnetic resonance images were acquired from 50 patients (33 PH, 17 non-PH). Contemporaneous right heart catheterization data were collected for all individuals. Outcome was defined by all-cause mortality and hospitalization for heart failure. RV endocardial borders were manually segmented, and three-dimensional surfaces reconstructed at end diastole and end systole. Registration and harmonic mapping were then used to create a quantitative correspondence between all RV surfaces. Proper orthogonal decomposition was performed to generate modes describing RV shape features. The first 15 modes captured over 98% of the total modal energy. Two shape modes, 8 (free wall expansion) and 13 (septal flattening), stood out as relating to PH state (mode 13: r = 0.424, p = 0.002; mode 8: r = 0.429, p = 0.002). Mode 13 was significantly correlated with outcome (r = 0.438, p = 0.001), more so than any hemodynamic variable. Shape analysis techniques can derive unique RV shape descriptors corresponding to specific, anatomically meaningful features. The modes quantify shape features that had been previously only qualitatively related to PH progression. Modes describing relevant RV features are shown to correlate with clinical measures of RV status, as well as outcomes. These new shape descriptors lay the groundwork for a noninvasive strategy for identification of failing RVs, beyond what is currently available to clinicians.
Databáze: OpenAIRE