Use of machine learning for wave separation analysis from non-invasive pressure measurement and its applicability in capturing aortic stiffening

Abstract

Background

Wave separation is considered as gold-standard method for investigating wave phenomena in the cardiovascular system. However, this method has not been fully translated into clinical practice due to its need for concurrent measurements of pressure and flow waveforms. In this study, we propose a new non-invasive pressure-only wave separation approach based on our recently introduced AI methodology (Aghilinejad et al. PMID: 37018682).

Methods

Collection of tonometry recordings of the carotid pressure and ultrasound measurements for the aortic flow waveforms is used from the Framingham Heart Study (2,640 individuals, 55% women) with mean (range) age of 66 (40-91) years. Reflection Magnitude (RM, defined as the ratio of peak backward to peak forward pressure) is used for assessing the approach.

Results

Method-derived estimate for the pressure-only RM is strongly correlated with the exact RM computed from pressure and flow measurements (r=0.82, no bias) (Fig. 1A and 1B). The normalized error between estimated and exact RM is 8.6% in the testing population (N=754) which is blinded to all stages of machine learning development. The impact of aortic stiffening (measured with carotid-femoral pulse wave velocity, adjusted for age and mean pulse pressure) is captured well using the pressure-only RM (Fig. 1C and 1D).

Conclusion

The proposed approach is a powerful tool to conduct wave separation analysis, and expand its clinical usage in affordable non-invasive devices such as wearable electronics.

Additional details