Development of a surrogate model for predicting atherosclerotic plaque progression based on agent based modeling data

Area of Science:

• Cardiovascular Research
• Computational Biology
• Biomedical Engineering

Background:

• Coronary atherosclerosis (CATS) is a leading global cause of death, characterized by plaque buildup in arteries.
• Computational modeling, particularly agent-based modeling (ABM), has advanced the simulation of plaque progression.
• There is a need for optimized predictive modeling resources, leading to the development of surrogate models.

Purpose of the Study:

• To develop a surrogate model for simulating atherosclerotic plaque progression.
• To utilize data from agent-based modeling (ABM) simulations for surrogate model training.
• To create a computationally efficient alternative to lengthy simulations.

Main Methods:

• Utilized a dataset comprising latin-hypercube sampling parameters and 15 patient-specific geometries with plaque progression data.
• Developed a deep learning-based surrogate model employing artificial neural networks (ANN).
• Benchmarked the surrogate model against the original ABM framework.

Main Results:

• The developed surrogate model achieved a high accuracy of 95.4% when benchmarked against the ABM model.
• Demonstrated the robustness and reliability of the artificial neural network framework for this application.
• Indicated the potential for accurate prediction of atherosclerotic plaque progression.

Conclusions:

• The high accuracy of the surrogate model supports its practical adoption.
• This framework enables the use of high-fidelity decision support systems for real-time prediction of atherosclerotic plaque progression.
• Facilitates faster and more efficient clinical decision-making in managing coronary artery disease.