Pulse transit time estimation of aortic pulse wave velocity and blood pressure using machine learning and simulated training data

Abstract

Recent developments in cardiovascular modelling allow us to simulate blood flow in an entire human body. Such model can also be used to create databases of virtual subjects, with sizes limited only by computational resources. In this work, we study if it is possible to estimate cardiovascular health indices using machine learning approaches. In particular, we carry out theoretical assessment of estimating aortic pulse wave velocity, diastolic and systolic blood pressure and stroke volume using pulse transit/arrival timings derived from photopletyshmography signals. For predictions, we train Gaussian process regression using a database of virtual subjects generated with a cardiovascular simulator. Simulated results provides theoretical assessment of accuracy for predictions of the health indices. For instance, aortic pulse wave velocity can be estimated with a high accuracy (r > 0.9) when photopletyshmography is measured from left carotid artery using a combination of foot-to-foot pulse transmit time and peak location derived for the predictions. Similar accuracy can be reached for diastolic blood pressure, but predictions of systolic blood pressure are less accurate (r > 0.75) and the stroke volume predictions are mostly contributed by heart rate.

Fig 1

(a) Illustration of human arterial system. The picture includes only a few largest arteries; see [19] for the complete set of arteries and veins used in the model. (b) Illustration of human heart including four chambers: left atrium (LA), left ventricle (LV), right atrium (RA) and right ventricle (RV). Left and right ventricular outflow track (lvot/rvot) are short 1D segments before the valves. Valves: tricuspid valve (TV), pulmonary valve (PV), mitral valve (MV) and aortic valve (AV). Picture by BruceBlaus (CC BY).

Fig 2

(a) Schematic of atrioventriclular (av) model. B is the Bernoulli valve resistance, R is the source resistance, L is the blood inertance and E is the elastance of the wall. The subscripts A and V refer to atrial and ventricular, respectively, and p_pc is the pericardiac pressure. (b) Freewall elastance E_fw for LA (blue) and LV (black). The figure includes four pulses. The duration of the pulse, the maximum elastance E_max and timing parameters τ₁ and τ₂ vary between pulses.

Fig 3

(a) Generic vascular bed model; (b) Hepatic vascular bed model with arterial and venous inlets; (c) Coronary vascular bed model with compartments representing subepicardial, midwall and subendocardial layers.

Fig 4. Two example pulses with the considered timings marked: the minimum/foot (PTT_ff; red cross), the maximum (PTT_p; green circle), the maximum of the first derivative (PTT_D; blue star), and dicrotic notch (DAT; magenta square).

Fig 5. Distributions of selected metrics for the virtual database (training set; after filtering): (a) heart rate (HR), (b) stroke volume (SV), (c) cardiac output (CO), (d) aortic PWV, (e) mean blood pressure (MBP), (f) pulse pressure (PP), (g) diastolic pressure (DPB), and (h) systolic blood pressure (SPB).

The means of the metrics are shown in the title.

Fig 6. Accuracy of the aortic PWV predictions using pulse transit time (PTT) measurements from left carotid artery (LCA).

Signals: heart rate (HR) and pulse transit times to the foot of signal (PTT_ff), peak of signal (PTT_p), the point of steepest raise (PTT_D), and the dicrotic notch (DAT).

Fig 7. Accuracy of the DBP predictions using pulse transit time (PTT) measurements from left carotid artery (LCA).

Signals: heart rate (HR) and pulse transit times to the foot of signal (PTT_ff), peak of signal (PTT_p), the point of steepest raise (PTT_D), and the dicrotic notch (DAT).

Fig 8. Accuracy of the SBP predictions using pulse transit time (PTT) measurements from left carotid artery (LCA).

Signals: heart rate (HR) and pulse transit times to the foot of signal (PTT_ff), peak of signal (PTT_p), the point of steepest raise (PTT_D), and the dicrotic notch (DAT).