Session: 07-02-01: Vibration and Acoustics in Biomedical Applications

Paper Number: 166766

Patient-Specific Myocardial Strain Estimation Using Optical Flow, Deep Learning, and Finite Element Modeling 

In 2011, United Nations member states committed to reducing premature mortality from non-communicable diseases (NCDs), including cardiovascular diseases (CVDs), by 25 percent by 2025. However, projections indicate that if current risk factor trends persist, the number of cardiovascular deaths is expected to rise from approximately 20.5 million in 2025 to 35.6 million by 2050. This underscores the critical need for effective interventions to address the growing burden of CVDs worldwide. Early detection is crucial for effective treatment. Traditional methods like measuring ejection fraction (EF) are commonly used to assess heart function. EF calculates the percentage of blood the heart pumps out with each beat. However, EF has limitations; it often detects heart problems only after significant damage has occurred. Furthermore, EF does not provide information about the localized motion of the myocardium or the strain distribution across the heart muscle. To overcome these limitations, recent computational advancements have enabled the estimation of myocardial strain directly from medical images. By using optical flow methods, deep learning-based motion estimation, and finite element modeling (FEM), researchers can now obtain more detailed and accurate strain measurements to capture the heart’s subtle mechanical changes with high spatial and temporal resolution. These approaches not only improve strain estimation but also allow for patient-specific modeling and this is what makes them superior to traditional methods. This method is particularly valuable in the early diagnosis of several cardiovascular diseases by detecting abnormal strain patterns before clinical symptoms arise. In heart failure with preserved ejection fraction (HFpEF), strain analysis reveals subtle myocardial dysfunction even when ejection fraction appears normal. Hypertrophic cardiomyopathy (HCM) and arrhythmogenic right ventricular cardiomyopathy (ARVC) present distinct strain patterns due to myocardial thickening or fibrofatty replacement, respectively. In myocardial infarction (MI), localized strain reductions indicate infarcted regions before significant loss of function. Diabetic cardiomyopathy is often detected by strain abnormalities that appear before noticeable systolic dysfunction. The computational methods provide a non-invasive approach to diagnose these diseases at an early stage. The assessment of myocardial strains (circumferential, longitudinal, and twist) provides comprehensive evaluation of the heart's function. Circumferential strain measures the contraction of myocardial fibers around the heart’s circumference and is particularly useful in detecting conditions that affect ventricular motions. Longitudinal strain evaluates the shortening and elongation of myocardial fibers along the long axis of the heart, mostly provides information about the function of the left ventricle (LV). Twist (shear) strain represents the rotational motion of the myocardium, which is important in evaluating cardiac pumping. A complete analysis of these strain components provides a more detailed understanding of cardiac mechanics than traditional metrics. Integrating advanced imaging techniques with computational modeling can enhance cardiac diagnostics. Optical flow methods in combine with deep learning algorithms, can track cardiac wall motion with high accuracy and provide essential data for constructing patient-specific FEM simulations. FEM models allow for biomechanical simulations that incorporate patient-specific tissue properties, heart geometry, and boundary conditions, that can enhance the accuracy of strain estimation. Our research aims to develop a framework that uses optical flow-based motion tracking, deep learning-based motion estimation and segmentation, and FEM simulations to assess myocardial strain. We will compare the strain calculations obtained from these three methods to determine their accuracy and clinical applicability. We expect to establish a computational approach for assessing myocardial strain. This research has the potential to improve early diagnosis to offer more precise assessments of cardiac function.

Presenting Author: Mohammadali Monfared Mississippi State University

Presenting Author Biography: Mohammadali Monfared is a PhD student of Biomedical Engineering at Mississippi State University. He holds a master's degree in Mechanical Engineering with a focus on Energy Conversion and Fluid Mechanics from Shiraz University, and a Bachelor of Science in Mechanical Engineering with a specialization in Fluid Mechanics from Persian Gulf University.

Authors:

Mohammadali Monfared Mississippi State University
Mohammad Muntasir Rahman Mississippi State University
Peshala Thibbotuwawa Gamage Florida Institute of Technology
Amirtahà Taebi Mississippi State University