Machine Learning based Estimation of Aortic Pressure Curves by Electrical Impedance Tomography.

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Medical Imaging

Background:

• Central aortic pressure is crucial for clinical practice but invasive or inaccurate noninvasive methods limit monitoring.
• There is a need for reliable, non-invasive methods for long-term central aortic pressure monitoring.

Purpose of the Study:

• To develop and evaluate a non-invasive method for predicting central aortic pressure curves using electrical impedance tomography (EIT).
• To assess the feasibility of using convolutional neural networks (CNNs) to estimate aortic pressure from EIT data.

Main Methods:

• EIT recordings were obtained from an in-vivo animal model.
• Simultaneous invasive central aortic pressure measurements were used for training.
• A CNN was trained to predict aortic pressure curves from EIT voltages, with parametric representations and hyperparameter tuning.

Main Results:

• The trained CNN successfully estimated aortic pressure curves from EIT data on unknown test data.
• The estimation was feasible, demonstrating the potential of EIT for pressure monitoring.
• Observed random offsets in the predicted pressure curves require further investigation.

Conclusions:

• Non-invasive estimation of central aortic pressure curves using EIT and CNNs is feasible.
• This approach offers a promising alternative to invasive or less accurate methods.
• Further research is needed to address and eliminate the observed random offsets for clinical application.