A novel method of artery stenosis diagnosis using transfer function and support vector machine based on transmission line model: A numerical simulation and validation study
Introduction
Artery stenosis is one kind of arterial disease in human cardiovascular system caused by the lesion of artery wall or congenital dysplasia disease, which usually occurs in the large and middle-sized arteries, such as carotid, cerebral, coronary and renal arteries [1], [2]. This disease directly affects the normal operation of relevant organs, causes organ insufficiency or degradation, and results in serious consequences: stroke, myocardial infarction, renal function loss, etc [3], [4]. Therefore, the early detection of arterial stenosis is very important for reducing the high morbidity and mortality caused by the kind disease.
The diagnosis technologies of artery stenosis mainly include digital subtraction angiography (DSA), transcranial Doppler (TD), computed tomography angiography (CTA), magnetic resonance angiography (MRA), etc [5], [6], [7]. Although these powerful technologies have advanced greatly, disadvantages still exist for each procedure. For example, DSA is an invasive method, though it is considered to be a gold standard; TD has low accuracy for the detection of stenosis of small arteries; CTA exposes patients to X-ray radiation and cannot completely prevent allergy to iodinated contrast agents; MRA is expensive. Moreover, pieces of equipment used in these technologies are mainly applied in hospitals and are not suitable for large area screenings. Therefore, a non-invasive, portable and simple technology has to be developed as a valuable complement to current methods, which is urgently needed for the large area screening of early artery stenosis [8], [9], [10].
The pulse-wave-based analysis method is a potential portable and noninvasive diagnosis technology of artery stenosis. Some useful parameters have been proposed based on this method, such as ankle brachial index (ABI) [11], pulse wave velocity (PWV) [12], transfer function (TF) [13], input impedance [14]. ABI is only used to detect the arterial stenosis of lower extremity, but not to detect the arterial stenosis of upper extremity, chest, abdomen, etc. Input impedance has been demonstrated numerically to be a valuable and useful parameter for the prediction of the artery stenosis in the human artery tree [15]. However, it is difficult to measure or calculate the input impedance of human arterial tree in clinic trials [14], [16]. The measurement or calculation of TF is easier than input impedance and also represents the mechanical and geometrical features of human arterial tree like input impedance. Because it is not affected by the blood pressure and flow waveform, TF has been successfully applied to noninvasively measure central blood pressure by using blood pressure waveform of radial or brachial artery, and achieved some good clinical results [17], [18].
Some researches have demonstrated there is a relation between TF and stenosis to some extent. Rajani [19] validated the central blood pressure and TF changed when some stenosis exist in human arterial tree. Gong [20] and Reinhard [21] evaluated the validity of the transfer function analysis in the assessment of human cerebral autoregulation in patients with carotid or basilar artery stenosis. Chao [22] conducted a study on comparisons of the baroreflex sensitivity and heart rate variability in patients with carotid stenosis and normal controls by using the method of TF analysis. Although a few valuable results and conclusions were obtained, the mathematical relation or model between TF and stenosis has not been established. In order to discover the potential of pulse-wave-based method for stenosis diagnosis, it is worthwhile to study the relation or modeling between of them.
In this paper, two aspects were mainly discussed: (1) A calculating method of TF was firstly proposed based on our previously presented transmission line model of 55 segment human arterial tree [23]. Then, one discussed the effects of arterial stenosis degree and position on the TF between ascending artery and carotid artery, radial artery and tibial artery. (2) Support vector machine (SVM) [24], [25], [26], [27], a machine learning method based on statistics, was introduced into the diagnosis of artery stenosis with TF as the feature vector of SVM model, and was validated by numerical simulation experiments. In the experiments, firstly, we built a database of samples using the transmission line model of human artery tree. Secondly, TF for each sample was calculated by a serial multiplication algorithm. Finally, two prediction models, one for stenosis existence, other for stenosis localization, were built by SVM and TF, and used to predict artery stenosis. The effects of the degree and position of artery stenosis on the prediction accuracy were discussed.
Section snippets
Arterial tree
The schematic diagram of the human arterial tree of this study is shown in Fig. 1. The original physiology data of the arterial tree was compiled by Noordergraaf et al. [28], and subsequently modified by Westerhof et al. [29], Avolio [30], Stergiopulos et al. [31], Wang and Parker [32], Liang et al. [33], and Alastruey and Parker et al. [34]. The arterial model of this study is based upon Stergiopuloss version of data that has 55 segments. The detail vascular dimensions and elastic constants
Transfer function of 55 segment human arterial tree without stenosis
In normal case, there is no stenosis in 55 segment human arterial tree. Transfer function between any two points in arterial tree can be calculated by using Eqs. (1), (2), (3), (4), (5), (6), (7), (8), (9), (10), (11), (12), (13). Generally, the TF between ascending aorta and other artery is the hot spot of research, such as radial artery, carotid artery and tibial artery. Fig. 2 showed the normal modulus and phase of TF in low frequency domain between the No. 1 ascending aorta to the No. 11
Conclusion
Based on the 55 segment human arterial tree, the effect of arterial stenosis on the modulus and phase of TF were discussed. In generally, the rule can be concluded as following four points: (1) for different arterial stenosis, the modulus of TF between any two points has a big difference, but the difference of phase is less obvious than the modulus; (2) when the stenosis occurs inside the path of two points for calculating the TF, there is a good correlation between the peak of modulus and the
Acknowledgements
This research is funded by the National Natural Science Foundation of China (grant no. 61501070) and Chongqing Natural Science Foundation (grant no. cstc2014jcyjA10040). This research is funded by Doctoral Scientific Research Foundation of Chongqing University of Technology (grant no. 2012ZD40).
References (47)
- et al.
Intracranial arterial stenosis
J. Stroke Cerebrovasc. Dis.
(2014) - et al.
Multiphase arterial spin labeling assessment of cerebral perfusion changes associated with middle cerebral artery stenosis
Acad. Radiol.
(2015) - et al.
Non-invasive evaluation of proximal vertebral artery stenosis using color Doppler sonography and CT angiography
J. Neuroradiol.
(2014) - et al.
Validation of a computer-aided diagnosis system for the automatic identification of carotid atherosclerosis
Ultrasound Med. Biol.
(2015) Computer-aided diagnosis in medical imaging: historical review, current status and future potential
Comput. Med. Imaging Graph.
(2007)- et al.
The use of an analog computer in a circulation model
Prog. Cardiovasc. Dis.
(1963) - et al.
Analog studies of the human systemic arterial tree
J. Biomech.
(1969) - et al.
Computer-simulation of arterial flow with applications to arterial and aortic stenoses
J. Biomech.
(1992) - et al.
Wave propagation in a model of the arterial circulation
J. Biomech.
(2004) - et al.
Application of knowledge discovery process on the prediction of stroke
Comput. Methods Prog. Biomed.
(2015)
Three way k-fold cross-validation of resource selection functions
Ecol. Modell.
Performance evaluation of classification algorithms by k-fold and leave-one-out cross validation
Pattern Recognit.
High frequency of intracranial arterial stenosis and cannabis use in ischaemic stroke in the young
Cerebrovasc. Dis.
The role of radiotherapy in the carotid stenosis
Ann. Ital. Chir.
Coronary artery stenosis in asymptomatic child after arterial switch operation: detection by transthoracic colour-flow Doppler echocardiography
Acta Paediatr.
Renal artery stenosis: comparative evaluation of gadolinium-enhanced MRA and DSA
Radiol. Med.
Correlation between brachial-ankle pulse wave velocity, carotid artery intima-media thickness, ankle-brachial index, and the severity of coronary lesions
Cell Biochem. Biophys.
Ankle brachial index is a valuable index of the severity of atherosclerotic renal artery stenosis
Scand. J. Urol. Nephrol.
The comparison of the ankle brachial index and pulse wave velocity between the patients with aortic stenosis and patients with bilateral lower extremity artery stenosis
Cardiology
Comparison of noninvasive assessments of central blood pressure using general transfer function and late systolic shoulder of the radial pressure wave
Am. J. Hypertens.
Recursive calculation and parameter analysis on input impedance of arterial tree based on electric network model
Yiyong Shengwu Lixue (J. Med. Biomech.)
Numerical simulation and validity of a novel method for the prediction of artery stenosis via input impedance and support vector machine
Biomed. Eng.: Appl. Basis Commun.
Input impedance of distributed arterial structures as used in investigations of underlying concepts in arterial haemodynamics
Med. Biol. Eng. Comput.
Cited by (16)
Peripheral artery disease diagnosis based on deep learning-enabled analysis of non-invasive arterial pulse waveforms
2024, Computers in Biology and MedicineAutodetect extracranial and intracranial artery stenosis by machine learning using ultrasound
2020, Computers in Biology and MedicineCitation Excerpt :To improve the accuracy of prediction, machine learning methods were employed. Literature suggests that the SVM classifier achieves superior and robust performance in stenosis detection [33–35]. The carotid Doppler parameters (extracranial data) and transcranial Doppler parameters (intracranial data) were used as the inputs of the SVM classifier.
A neural network approach to classify carotid disorders from Heart Rate Variability analysis
2019, Computers in Biology and MedicineCitation Excerpt :An ANN was also used by Samiappan et al. [27] to classify carotid abnormalities in 361 US carotid artery images selected from a private database. A diagnosis model of artery stenosis was built, instead, by using the SVM and transfer function (TF), important parameters for the analysis and understanding of the hemodynamics of the carotid arteries, in a study presented in Ref. [28]. Table 1 summarizes the different approaches outlined so far.
Clinically-Driven Virtual Patient Cohorts Generation: An Application to Aorta
2021, Frontiers in PhysiologyGeometric Mean Maximum FSVMI Model and Its Application in Carotid Artery Stenosis Risk Prediction
2021, Chinese Journal of ElectronicsIntelligent Oscillometric System for Automatic Detection of Peripheral Arterial Disease
2021, IEEE Journal of Biomedical and Health Informatics