Mapping Vagus Nerve Stimulation Parameters to Cardiac Physiology using Long Short-term Memory Network

Area of Science:

• Biomedical Engineering
• Computational Physiology
• Neuroscience

Background:

• Vagus nerve stimulation (VNS) shows therapeutic potential but faces challenges in automated closed-loop control.
• Developing predictive models is crucial for effective VNS implementation in biological systems.

Purpose of the Study:

• To propose and validate a data-driven approach for predicting the physiological impact of VNS.
• To demonstrate the utility of this approach in the cardiac system for closed-loop control applications.

Main Methods:

• Utilized a synthetic dataset generated from a physiological rat heart model.
• Trained and evaluated various neural network models, focusing on Long Short-Term Memory (LSTM) architecture.
• Assessed the model's ability to map VNS parameters to physiological responses.

Main Results:

• The LSTM neural network architecture achieved the best performance on the test dataset.
• The model successfully predicted the impact of VNS on heart rate and mean arterial blood pressure.
• Demonstrated a data-driven method to create a required physiological model for closed-loop control.

Conclusions:

• A data-driven approach using LSTM networks can effectively predict VNS effects on cardiac function.
• This method addresses the need for physiological models in automated closed-loop VNS control.
• The findings support the advancement of VNS as a therapeutic strategy through improved control mechanisms.