Translating Mechanistic Insights Into Action and Revealing New Pathways: Machine Learning Approaches in Heart Failure With Preserved Ejection Fraction

Figure 1. Machine learning approaches for integrating pathophysiologic mechanisms in HFpEF to advance precision medicine. Left, key mechanisms of cardiac dysfunction in HFpEF; Middle, highlighting how machine learning can integrate these diverse data sources; and Right, multimodal integration with the goal of advancing precision medicine. HFpEF indicates heart failure with preserved ejection fraction.

Figure 2. Three‐dimensional finite element model reconstruction to estimate myocardial stiffness using cardiac magnetic resonance imaging data. FE indicates finite element; MRI, magnetic resonance imaging

Figure 3. Estimation of myocardial stiffness using a machine learning approach. A, 3‐dimensional reconstruction of human cardiac anatomy at end‐diastole based on CMR images. This geometric information is then used in simulations to estimate mechanical behavior (motion) as in (B), which visualizes the displacement of the cardiac tissue. These data are then used by the ML model to predict myocardial stiffness. Figure adapted from A machine learning model to estimate myocardial stiffness from EDPVR by Babaei et al. 18 , published in Scientific Reports under a CC BY license. © [2022]. CMR indicates cardiac magnetic resonance; and ML, machine learning.

Figure 4. Diffusion tensor imaging scan and fiber angle visualization. An example of a cardiac magnetic resonance diffusion tensor imaging short‐axis slice from a control (top left) and patient with heart failure with preserved ejection fraction (bottom left) with corresponding 3‐dimensional fiber orientation angle visualization (right). HFpEF indicates heart failure with preserved ejection fraction.

Figure 5. Multifidelity physics‐informed neural networks incorporating demographic, clinical, echocardiographic, and magnetic resonance imaging variables. Layer 1 (left) depicts demographic and clinical variables; layers 2 and 3 (right) illustrate integration of advanced imaging variables. CMR indicates cardiac magnetic resonance.