Machine learning and physical based modeling for cardiac hypertrophy.

Area of Science:

• Cardiology
• Biomedical Engineering
• Computational Biology

Background:

• Predicting left ventricular remodeling and expansion in patients is clinically significant but challenging.
• Cardiac hypertrophy, a condition of left ventricular enlargement, requires accurate long-term monitoring.

Purpose of the Study:

• To develop and compare machine learning (ML) and physics-based (finite element) models for predicting cardiac hypertrophy.
• To assess the accuracy and speed of ML and finite element models in forecasting disease evolution.

Main Methods:

• Trained ML models (random forests, gradient boosting, neural networks) using patient medical history and cardiac health data.
• Developed a physics-based finite element model simulating cardiac hypertrophy development.
• Collected multi-patient data for model training and validation.

Main Results:

• Both ML and finite element models accurately forecasted hypertrophy evolution over six years, showing similar outcomes.
• The finite element model demonstrated higher accuracy due to its basis in physical laws.
• The ML model exhibited significantly faster computation times.

Conclusions:

• ML models offer a faster, potentially clinically applicable approach for monitoring cardiac hypertrophy.
• Finite element models provide greater accuracy but are computationally intensive.
• Integrating finite element simulation data into ML models can enhance speed and accuracy for clinical use.