Filter-type neural network-based counter-pulsation control in pulsatile ECMO: improving heartbeat-pulse...

Area of Science:

• Biomedical Engineering
• Cardiovascular Technology
• Artificial Intelligence in Medicine

Background:

• Pulsatile extracorporeal membrane oxygenation (p-ECMO) aims to reduce risks associated with conventional ECMO.
• Accurate heartbeat detection is crucial for p-ECMO counter-pulsation (CP), especially when electrocardiograms (ECGs) are unavailable.
• Blood pressure (BP) waveforms contain vital pulse information for CP control.

Purpose of the Study:

• To develop and evaluate a novel algorithm for heartbeat detection in BP waveforms for p-ECMO CP control.
• To implement a control system that uses detected heart rate (HR) to synchronize p-ECMO pulses with heartbeats.
• To assess the efficacy of the CP control system under varying HR conditions using a mock circulation system.

Main Methods:

• A cumulative algorithm utilizing filter-type neural networks was designed to differentiate heartbeats from noise in BP data.
• A control system was implemented to detect HR and regulate the timing of p-ECMO pulses for CP.
• A mock circulation system with a heart model was used to simulate human BP waveforms for testing.

Main Results:

• The algorithm successfully maintained CP during constant HR conditions.
• A 0.48-second delay in HR detection to CP control impacted performance during sudden HR increases.
• CP control success rate decreased to 78.62% with HR variations of ±5 bpm, significantly outperforming the 25.75% success rate without control.

Conclusions:

• The developed neural network algorithm shows promise for enabling CP in p-ECMO systems by detecting heartbeats from BP data.
• The system's performance is sensitive to HR fluctuations due to detection and control delays.
• Further optimization is needed to address the delay issue for robust CP control in dynamic physiological conditions.