Early thrombus detection in ECMO with optimized impedance measurements: A simulative study

Area of Science:

• Biomedical Engineering
• Computational Modeling
• Medical Device Technology

Background:

• Extracorporeal oxygenation is vital for severe cardiac/respiratory failure.
• Thrombus formation in oxygenators reduces efficiency and poses risks like embolism.
• Current thrombus detection methods lack accuracy and timeliness.

Purpose of the Study:

• To develop a computational bioimpedance approach for early thrombus detection in oxygenators.
• To optimize detection sensitivity and accuracy using advanced modeling and machine learning.
• To ensure the method preserves oxygenator functionality.

Main Methods:

• Developed a finite element model of an oxygenator for bioimpedance simulation.
• Utilized neural networks to optimize electrode configurations and measurement patterns.
• Trained a second neural network on simulated data for thrombus classification.

Main Results:

• Optimized electrode placements significantly enhanced sensitivity to conductivity changes.
• The classification neural network achieved an F1-score exceeding 94% for thrombus detection.
• Simulations confirmed the feasibility and improved accuracy of the computational bioimpedance method.

Conclusions:

• Computational bioimpedance offers a robust framework for automated thrombus detection.
• Neural network optimization is key to enhancing sensitivity and accuracy.
• This approach promises improved safety and efficiency in extracorporeal oxygenation systems.