Use of AI and a non-invasively obtained estimate of the total blood pressure waveform for cardiovascular disorder screening

Abstract

Accurate assessment of cardiovascular pathologies is essential for routine and specific patient diagnoses. Because heart disease is prevalent world-wide, an inexpensive, reliable, and non-invasive screening method that evaluates blood pressure, arterial stiffness, atrial fibrillation and other very specific pathologies in the cardiovascular system would improve diagnoses and potentially save many lives. We propose that technology is available for the diagnostic interpretation of patient data acquired from simple oscillometric blood pressure cuffs. Specifically, a model-based identification algorithm has been used successfully to estimate the coefficients in a Fourier series equation for the waveform while simultaneously estimating the coefficients in an empirical model for arterial stiffness. Thus, the continuous waveform with all its peaks and valleys including the dicrotic notch, can be displayed on an oscilloscope and/or plotted on a graph for visual examination. More importantly, the mathematical format for the waveform and arterial stiffness can be accessed by machine learning algorithms to diagnose and identify appropriate treatments. Research has revealed correlations of the waveform features and arterial properties with cardiovascular disorders. We propose the creation of a brachial artery database with cuff pressure time histories for specific cuff designs and patient medical histories. Machine learning algorithms can use this database to enhance the accuracy of diagnoses and treatments as well as be available for testing and validating more robust identification algorithms proposed in the future.