Transient wall shear stress estimation in coronary bifurcations using convolutional neural networks

Area of Science:

• Cardiovascular research
• Biomedical engineering
• Artificial intelligence in medicine

Background:

• Blood flow shear stress is linked to coronary artery disease development.
• Traditional Computational Fluid Dynamics (CFD) for calculating Wall Shear Stress (WSS) is slow and computationally expensive.
• CFD is not suitable for large-scale clinical use.

Purpose of the Study:

• To develop a deep learning model for predicting luminal WSS in coronary bifurcations.
• To offer a computationally efficient alternative to CFD for haemodynamic analysis.
• To enable large-scale population studies and potential clinical integration.

Main Methods:

• Deep learning techniques were used to predict WSS magnitude.
• The model utilized steady-state solutions, vessel geometry, and global features.
• Training involved 101 patient-specific and 2626 synthetic left main bifurcation models.

Main Results:

• The deep learning model achieved high-fidelity predictions with less than 5% deviation from CFD values.
• The model was significantly faster, with computation times under 2 minutes compared to 3 hours for transient CFD.
• This demonstrates a substantial reduction in computational cost.

Conclusions:

• Deep learning provides a rapid and accurate method for calculating coronary haemodynamic metrics.
• This approach can significantly reduce computational costs for population studies.
• The method holds potential for future integration into clinical settings for coronary artery disease risk assessment.