Artificial Intelligence Estimation of Carotid-Femoral Pulse Wave Velocity using Carotid Waveform

Abstract

In this article, we offer an artificial intelligence method to estimate the carotid-femoral Pulse Wave Velocity (PWV) non-invasively from one uncalibrated carotid waveform measured by tonometry and few routine clinical variables. Since the signal processing inputs to this machine learning algorithm are sensor agnostic, the presented method can accompany any medical instrument that provides a calibrated or uncalibrated carotid pressure waveform. Our results show that, for an unseen hold back test set population in the age range of 20 to 69, our model can estimate PWV with a Root-Mean-Square Error (RMSE) of 1.12 m/sec compared to the reference method. The results convey the fact that this model is a reliable surrogate of PWV. Our study also showed that estimated PWV was significantly associated with an increased risk of CVDs.Cardiovascular diseases (CVDs) and stroke are among the major causes of death in the United States and the total cost related to them was more than $316 billion in 2011-2012 1,2 . New cardiovascular monitoring methods are urgently needed in order to limit the growing burden of CVDs. Arterial stiffening is one of the risk factors for CVDs 3,4 , which can be assessed non-invasively by calculating the carotid to femoral PWV 5 . This parameter is a gold standard of arterial stiffness, the rate at which pressure waves move down the aortic vessel 6 . Increased arterial stiffness is related to an increased risk of cardiovascular events; therefore, it has become an independent marker for CVDs 6,7 . Because of its clinical significance, there has been a surge in addressing arterial stiffness and PWV 8 . Arterial stiffness and its surrogates such as PWV have been suggested as one of the risk factors along with other biomarkers such as high cholesterol, diabetes, and left ventricular hypertrophy when cardiovascular risk is being evaluated 8 . Past studies have shown a strong correlation between PWV and the presence of CVDs 9-14 .Although carotid-femoral PWV measurement is non-invasive, this process is intrusive as it requires the waveform collection from inguinal region. Obtaining accurate carotid-femoral PWV measurements often requires a well-trained staff within a clinical setting 15 . The need of the medical community is an easy-to-use and non-intrusive method to measure carotid-femoral PWV with acceptable accuracy and precision; see ref. 16 .At the same time, recent advances in the field of artificial intelligence have opened up new areas and methods in creating novel modeling and predictive methods for clinical use 17 . The model and analysis in this paper are in accord to this path of introducing artificial intelligence to the field of medical sciences.In this study, a novel, easy-to-use, and non-invasive approach to estimate carotid-femoral PWV, from a single carotid waveform measurement, is explored. This method is based on the newly developed Intrinsic Frequency (IF) algorithm 18,19 . IF method solely needs one uncalibrated trace of a carotid, or aortic, pressure waveform. Our method takes an un...

Cardiac Triangle Mapping: A New Systems Approach for Noninvasive Evaluation of Left Ventricular End Diastolic Pressure

“…Recently, we introduced the intrinsic frequency (IF) concept that reveals clinically relevant information about the dynamics of the left ventricle (LV), arterial system, and their interactions [7][8][9][10]. IF frequencies are operating frequencies based on the Sparse Time-Frequency Representation (STFR) [11], treating the LV combined with the aorta and the remaining peripheral arteries as a coupled dynamical system (heart + aortic tree), which is decoupled upon closure of the aortic valve [8,9,12]. Utilizing the IF method, one can extract two IF frequencies, ω 1 and ω 2 (along with other independent variables [9]), from a single arterial blood pressure waveform.…”

“…The first IF, ω 1 , describes the dynamics of the systolic phase of the cardiac cycle where the LV and aorta (vasculature) are a coupled dynamic system [7][8][9]. The second IF, ω 2 , is dominated by the dynamics of the vasculature [9,10,12].…”

“…Previous studies have indicated that ω 1 is mainly determined by the LV contractility while ω 2 is mostly determined by vascular function (e.g., arterial stiffness, arterial wave dynamics, and afterload) [9,10,12,14]. In a blind clinical study [8], it was shown LV ejection fraction (LVEF, a surrogate of LV contractility) derived using IF applied to carotid waveforms extracted using an iPhone [8] were similar to those measured from cardiac magnetic resonance imaging (CMRI) (r = 0.74, p < 0.00001).…”

“…Using the CMRI as the gold standard, the results showed that the detection of cardiac systolic dysfunction (as measured by LVEF) via the noninvasive LVEF-IF method was more accurate and sensitive than 2D echocardiography [13]. Using Framingham Heart Study data, we have recently demonstrated that central arterial stiffness (carotid-femoral pulse wave velocity (PWV)) can be computed using IF of a single noninvasively measured carotid pressure waveform (without any need for ECG measurement and/or femoral tonometry) [12]. The result showed that estimated PWV by IF method is equivalent to PWV measurements obtained by direct methods in predicting the risk for CVD [12].…”

“…Using Framingham Heart Study data, we have recently demonstrated that central arterial stiffness (carotid-femoral pulse wave velocity (PWV)) can be computed using IF of a single noninvasively measured carotid pressure waveform (without any need for ECG measurement and/or femoral tonometry) [12]. The result showed that estimated PWV by IF method is equivalent to PWV measurements obtained by direct methods in predicting the risk for CVD [12]. These studies point to the importance of the IF method as a powerful tool for noninvasive ventricular performance (contractility) and vascular function (afterload) evaluation [8,9,12,13].…”