Efficient approximation of cardiac mechanics through reduced-order modeling with deep learning-based operator...

Area of Science:

• Computational mechanics
• Biomedical engineering
• Artificial intelligence in healthcare

Background:

• High-fidelity cardiac mechanics simulations (full-order models) are computationally expensive, limiting clinical use.
• Accurate simulations require fine spatio-temporal discretizations, leading to hours of computation even for a few heartbeats.
• Patient-specific cardiac models require parameter calibration for virtual scenario exploration, further increasing computational demands.

Purpose of the Study:

• To develop a computationally efficient method for patient-specific cardiac mechanics simulations.
• To reduce the computational time of full-order models (FOMs) for clinical translation.
• To enable advanced analyses like uncertainty quantification in cardiac mechanics.

Main Methods:

• A reduced basis method combined with deep neural networks for operator approximation (Deep-HyROMnet).
• Projection-based Proper Orthogonal Decomposition-Galerkin method integrated with deep learning.
• Coupling a 3D cardiac tissue mechanics model with a 0D blood circulation model and a parameter-dependent surrogate for active force generation.

Main Results:

• Deep-HyROMnet achieves orders of magnitude computational speed-up compared to classical projection-based reduced-order models (ROMs).
• The method provides highly accurate approximations for patient-specific cardiac mechanics, including the complete cardiac cycle.
• Accurate pressure-volume loops were reproduced for both physiological and pathological cases.
• Forward uncertainty quantification analysis, previously unaffordable with FOMs, became feasible.

Conclusions:

• The Deep-HyROMnet technique offers a significant advancement in computational cardiac mechanics.
• This approach accelerates patient-specific simulations, paving the way for clinical integration.
• Enables complex analyses like uncertainty quantification for improved understanding of cardiac function and disease.