A spectral machine learning approach to derive central aortic pressure waveforms from a brachial cuff

Area of Science:

• Cardiovascular Physiology
• Biomedical Engineering
• Machine Learning in Medicine

Background:

• Central aortic pulse waveforms provide critical cardiovascular insights but are difficult to obtain noninvasively.
• Current methods for central pressure measurement are often invasive or lack precision.

Purpose of the Study:

• To develop and validate a noninvasive method for reconstructing central aortic pulse waveforms using brachial artery measurements.
• To assess the clinical performance of a spectral machine learning model for predicting central hemodynamics.

Main Methods:

• A laboratory-developed cuff device acquired high-resolution brachial pulse waveforms.
• A spectral machine learning model nonlinearly mapped brachial wave components to the aortic site.
• Simultaneous invasive aortic catheter and brachial cuff waveforms were collected from 115 subjects.

Main Results:

• The noninvasive method reconstructed aortic waveforms with high fidelity (11.3% error).
• It accurately captured dynamic systolic blood pressure oscillations (r=0.76) and correlated well with invasive measurements (systolic BP R²=0.83, diastolic BP R²=0.58).
• Shape-based features, including pressure-time integrals and subendocardial viability ratio, showed strong correlations (r=0.86-0.95).

Conclusions:

• The proposed wave-based approach effectively predicts central aortic waveform morphology noninvasively.
• This technique holds significant promise for advancing noninvasive cardiovascular diagnostics in clinical cardiology.