Search

This study presents an application of machine learning (ML) methods for detecting the presence of stenoses and aneurysms in the human arterial system. Four major forms of arterial disease -- carotid artery stenosis (CAS), subclavian artery stenosis (SAC), peripheral arterial disease (PAD), and abdominal aortic aneurysms (AAA) -- are considered. The ML methods are trained and tested on a physiologically realistic virtual patient database (VPD) containing 28,868 healthy subjects, which is adapted from the authors previous work and augmented to include the four disease forms. Six ML methods -- Naive Bayes, Logistic Regression, Support Vector Machine, Multi-layer Perceptron, Random Forests, and Gradient Boosting -- are compared with respect to classification accuracies and it is found that the tree-based methods of Random Forest and Gradient Boosting outperform other approaches. The performance of ML methods is quantified through the F1 score and computation of sensitivities and specificities. When using all the six measurements, it is found that maximum F1 scores larger than 0.9 are achieved for CAS and PAD, larger than 0.85 for SAS, and larger than 0.98 for both low- and high-severity AAAs. Corresponding sensitivities and specificities are larger than 90% for CAS and PAD, larger than 85% for SAS, and larger than 98% for both low- and high-severity AAAs. When reducing the number of measurements, it is found that the performance is degraded by less than 5% when three measurements are used, and less than 10% when only two measurements are used for classification. For AAA, it is shown that F1 scores larger than 0.85 and corresponding sensitivities and specificities larger than 85% are achievable when using only a single measurement. The results are encouraging to pursue AAA monitoring and screening through wearable devices which can reliably measure pressure or flow-rates, Comment: Changelog: Added a bullet point in the discussion (end of section 3.3)

This study creates a physiologically realistic virtual patient database (VPD), representing the human arterial system, for the primary purpose of studying the affects of arterial disease on haemodynamics. A low dimensional representation of an anatomically detailed arterial network is outlined, and a physiologically realistic posterior distribution for its parameters is constructed through a Bayesian approach. This approach combines both physiological/geometrical constraints and the available measurements reported in the literature. A key contribution of this work is to present a framework for including all such available information for the creation of virtual patients (VPs). The Markov Chain Monte Carlo (MCMC) method is used to sample random VPs from this posterior distribution, and the pressure and flow-rate profiles associated with the VPs are computed through a model of pulse wave propagation. This combination of the arterial network parameters (representing the VPs) and the haemodynamics waveforms of pressure and flow-rates at various locations (representing functional response of the VPs) makes up the VPD. While 75,000 VPs are sampled from the posterior distribution, 10,000 are discarded as the initial burn-in period. A further 12,857 VPs are subsequently removed due to the presence of negative average flow-rate. Due to an undesirable behaviour observed in some VPs -- asymmetric under- and over-damped pressure and flow-rate profiles in the left and right sides of the arterial system -- a filter is proposed for their removal. The final VPD has 28,868 subjects. It is shown that the methodology is appropriate by comparing the VPD statistics to those reported in literature across real populations. A good agreement between the two is found while respecting physiological/geometrical constraints. The pre-filter database is made available at https://doi.org/10.5281/zenodo.4549764.

This proof of concept (PoC) assesses the ability of machine learning (ML) classifiers to predict the presence of a stenosis in a three vessel arterial system consisting of the abdominal aorta bifurcating into the two common iliacs. A virtual patient database (VPD) is created using one-dimensional pulse wave propagation model of haemodynamics. Four different machine learning (ML) methods are used to train and test a series of classifiers -- both binary and multiclass -- to distinguish between healthy and unhealthy virtual patients (VPs) using different combinations of pressure and flow-rate measurements. It is found that the ML classifiers achieve specificities larger than 80% and sensitivities ranging from 50-75%. The most balanced classifier also achieves an area under the receiver operative characteristic curve of 0.75, outperforming approximately 20 methods used in clinical practice, and thus placing the method as moderately accurate. Other important observations from this study are that: i) few measurements can provide similar classification accuracies compared to the case when more/all the measurements are used; ii) some measurements are more informative than others for classification; and iii) a modification of standard methods can result in detection of not only the presence of stenosis, but also the stenosed vessel.