Hemodynamic optimization of cardiac resynchronization therapy by heart sound sensor in pulse generator

Abstract

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): EU-HORIZON 2020-International Training Network H2020-MSCA-ITN-2017. Background Continuous ambulatory AV-optimization for Cardiac Resynchronization Therapy (CRT) is mainly performed by electrical means. Objective Develop an estimation model of cardiac function that uses feedback from a piezoelectric transducer embedded in an internal pulse generator to guide CRT optimization. Methods During implantation procedures measurements were performed of left ventricular pressure (LVP) and heart sound waveforms, the latter using the piezoelectric alarm transducer of the device in 22 CRT patients at atrioventricular (AV)-delays of 60-330 ms. Post-hoc, machine learning (ML) was employed to produce models that use heart-sound based features to estimate cardiac function in terms of maximum LVP (LVPmax) and the maximum positive derivative of LVP (LVdP/dtmax). Polynomial curve fitting was used to compare estimated and measured AV-delays with highest LVPmax and LVdP/dtmax. Results The figure displays the concept of the study and an example of the measured and estimated optimization curve. With a dataset of ~30,000 beats, ML provided three most suitable features. Using LVPmax, ML-based estimated optimal AV-delays were not significantly different from measured ones (-5.6 ± 17.1ms for LVPmax and +5.1 ± 46.7ms for LVdP/dtmax). The difference in function between the estimated and measured optimal AV-delays was small and not statically significant (1 ± 3mmHg for LVPmax and 9 ± 57mmHg/s for LVdP/dtmax). Conclusion Heart sound sensors in devices in combination with ML provide a reliable assessment of absolute LVPmax and LVdP/dtmax and may be used for continuous optimization of AV-delays in CRT patients.