Estimation of wave reflection in aorta from radial pulse waveform by artificial neural network: a numerical study

Personalized aortic pressure waveform estimation from brachial pressure waveform using an adaptive transfer function

BACKGROUND AND OBJECTIVE

The aortic pressure waveform (APW) provides reliable information for the diagnosis of cardiovascular disease. APW is often measured using a generalized transfer function (GTF) applied to the peripheral pressure waveform acquired noninvasively, to avoid the significant risks of invasive APW acquisition. However, the GTF ignores various physiological conditions, which affects the accuracy of the estimated APW. To solve this problem, this study utilized an adaptive transfer function (ATF) combined with a tube-load model to achieve personalized and accurate estimation of APW from the brachial pressure waveform (BPW).

METHODS

The proposed method was validated using APWs and BPWs from 34 patients. The ATF was defined using a tube-load model in which pulse transit time and reflection coefficients were determined from, respectively, the diastolic-exponential-pressure-decay of the APW and a piece-wise constant approximation. The root-mean-square-error of overall morphology, mean absolute errors of common hemodynamic indices (systolic blood pressure, diastolic blood pressure and pulse pressure) were used to evaluate the ATF.

RESULTS

The proposed ATF performed better in estimating diastolic blood pressure and pulse pressure (1.63 versus 1.94 mmHg, and 2.37 versus 3.10 mmHg, respectively, both P < 0.10), and produced similar errors in overall morphology and systolic blood pressure (3.91 versus 4.24 mmHg, and 2.83 versus 2.91 mmHg, respectively, both P > 0.10) compared to GTF.

CONCLUSION

Unlike the GTF which uses fixed parameters trained on existing clinical datasets, the proposed method can achieve personalized estimation of APW. Hence, it provides accurate pulsatile hemodynamic measures for the evaluation of cardiovascular function.

Estimation of cardiac stroke volume from radial pulse waveform by artificial neural network

BACKGROUND AND OBJECTIVES

Stroke volume (SV) and cardiac output (CO) are the key indicators for the evaluation of cardiac function and hemodynamic status during the perioperative period, which are very important in the detection and treatment of cardiovascular diseases. Traditional CO and SV measurement methods have problems such as complex operation, low precision and poor generalization ability.

METHODS

In this paper, a method for estimating stroke volume based on cascade artificial neural network (ANN) and time domain features of radial pulse waveform (SV_ANN) was proposed. The simulation datasets of 4000 radial pulse waveforms and stroke volume (SV_meas) were generated by a 55 segment transmission line model of the human systemic vasculature and a recursive algorithm. The ANN was trained and tested by 10-fold cross-validation, and compared with 12 traditional models.

RESULTS

Experimental results showed that the Pearson correlation coefficients and mean difference between SV_ANN and SV_meas (R=0.95, mean standard deviation (SD) = 0.00 ± 6.45) were better than the best results of the 12 traditional models. Moreover, as increasing the number of training samples, the performance improvement of the ANN (R=0.94(Δ + 0.04), mean ± SD = 0.00 ± 6.38(Δ± 2.02)) was better than the other best model, namely, multiple linear regression model (MLR) (R=0.93(Δ + 0.03), mean ± SD = 0.00 ± 6.99(Δ± 1.50)).

CONCLUSIONS

A method is proposed to estimate cardiac stroke volume by the ANN with time domain features of radial pulse wave. It avoids the complicated modeling process based on hemodynamics within traditional models, improves the estimation accuracy of SV, and has a good generalization ability.

A method of parameter estimation for cardiovascular hemodynamics based on deep learning and its application to personalize a reduced-order model

Precise model personalization is a key step towards the application of cardiovascular physical models. In this manuscript, we propose to use deep learning (DL) to solve the parameter estimation problem in cardiovascular hemodynamics. Based on the convolutional neural network (CNN) and fully connected neural network (FCNN), a multi-input deep neural network (DNN) model is developed to map the nonlinear relationship between measurements and the parameters to be estimated. In this model, two separate network structures are designed to extract the features of two types of measurement data, including pressure waveforms and a vector composed of heart rate (HR) and pulse transit time (PTT), and a shared structure is used to extract their combined dependencies on the parameters. Besides, we try to use the transfer learning (TL) technology to further strengthen the personalized characteristics of a trained-well network. For assessing the proposed method, we conducted the parameter estimation using synthetic data and in vitro data respectively, and in the test with synthetic data, we evaluated the performance of the TL algorithm through two individuals with different characteristics. A series of estimation results show that the estimated parameters are in good agreement with the true values. Furthermore, it is also found that the estimation accuracy can be significantly improved by a multicycle combination strategy. Therefore, we think that the proposed method has the potential to be used for parameter estimation in cardiovascular hemodynamics, which can provide an immediate, accurate, and sustainable personalization process, and deserves more attention in the future.

Estimation of aortic pulse wave velocity based on waveform decomposition of central aortic pressure waveform

Objective.Aortic stiffness is associated with risk of cardiovascular events. Carotid-femoral pulse wave velocity (cfPWV) is the current noninvasive gold standard for assessing aortic stiffness. However, the cfPWV measurement is challenging, requiring simultaneous signals at the carotid and femoral sites.Approach.In this study, the aortic PWV is estimated using a single radial pressure waveform and compared with cfPWV. 111 subjects' aortic PWVs are estimated from the decomposition of the derived central aortic pressure waveform based on three types of reconstructed flow waveform: the peak of triangular flow waveform based on 30% ejection time (Q_30%tri), the peak of triangular flow waveform based on inflection point (Q_tri), and averaged flow waveform (Q_avg). The central aortic pressure waveform is derived from a radial pressure waveform via a validated transfer function.Main results.TheQ_avgis used for estimating aortic PWV without the determination of the peak point of the triangular flow waveforms. The estimated aortic PWV shows good agreement with cfPWV. The mean difference ± SD is 0.29 ± 1.50 m s^-1(r² = 0.29,p< 0.001) for theQ_30%tri; 0.27 ± 1.40 m s^-1(r² = 0.38,p < 0.001) for theQ_tri; 0.23 ± 1.39 m s^-1(r² = 0.40,p < 0.001) for theQ_avg. The correlation between estimated aortic PWV based onQ_30%triand measured cfPWV is weak. The results ofQ_triandQ_avgshow no obvious difference.Significance.The proposed method can be used as a less complex way than conventional measurement of cfPWV to further assess arterial stiffness and predict cardiovascular risks or events.

The composition of vulnerable plaque and its effect on arterial waveforms

Carotid plaque composition is a key factor of plaque stability and it carries significant prognostic information. The carotid unstable plaques are characterized by a thin fibrous cap (FC) ≤65μm with large lipid core (LC), while stable plaques have a thicker FC and less LC. Identifying the percentage of plaque compositions could help surgeons to make a precise decision for their patients' treatment protocol. This study aims to distinguish between stable and unstable plaque by defining the relationship between plaque composition and arterial waveform non-invasively. An in-vitro arterial system, composed of a Harvard pulsatile flow pump and artificial circulation system, was used to investigate the effect of the plaque compositions on the pulsatile arterial waveforms. Five types of arterial plaques, composed of the LC, FC, Collagen (Col) and Calcium (Ca), were implemented into the artificial carotid artery to represent the diseased arterial system with 30% of blockage. The pulsatile pressure, velocity and arterial wall movement were measured simultaneously at the site proximal to the plaque. Non-invasive wave intensity analysis (Non-WIA) was used to separate the waves into forward and backward components. The correlation between the plaque compositions and the reflected waveforms was quantitatively analysed. The experimental results indicate that the reflected waveforms are strongly correlated with the plaque compositions, where the percentages of the Col are linearly correlated with the amplitude of the backward diameter (correlation coefficient, r = 0.74) and the lipid content has a strong negative correlation with the backward diameter (r = 0.82). A slight weak correlation exists between the reflected waveform and the percentage of Ca. The strong correlation between the compositions of the plaques with the backward waveforms observed in this study demonstrates that the components of the arterial plaques could be distinguished by the arterial waveforms. This finding might lead to a potential novel non-invasive clinical tool to determine the composition of the plaques and distinguish between stable and vulnerable arterial plaques at the early stage.