Sanjay Pant

Yue Mei, Daniel E. Hurtado, Sanjay Pant et al.

Finite elasticity problems commonly include material and geometric nonlinearities and are solved using various numerical methods. However, for highly nonlinear problems, achieving convergence is relatively difficult and requires small load step sizes. In this work, we present a new method to transform the discretized governing equations so that the transformed problem has significantly reduced nonlinearity and, therefore, Newton solvers exhibit improved convergence properties. We study exponential-type nonlinearity in soft tissues and geometric nonlinearity in compression, and propose novel formulations for the two problems. We test the new formulations in several numerical examples and show significant reduction in iterations required for convergence, especially at large load steps. Notably, the proposed formulation is capable of yielding convergent solution even when 10 to 100 times larger load steps are applied. The proposed framework is generic and can be applied to other types of nonlinearities as well.

Ankush Aggarwal, Sanjay Pant

Finding roots of equations is at the heart of most computational science. A well-known and widely used iterative algorithm is the Newton's method. However, its convergence depends heavily on the initial guess, with poor choices often leading to slow convergence or even divergence. In this paper, we present a new class of methods that improve upon the classical Newton's method. The key idea behind the new approach is to develop a relatively simple multiplicative transformation of the original equations, which leads to a significant reduction in nonlinearities, thereby alleviating the limitations of the Newton's method. Based on this idea, we propose two novel classes of methods and present their application to several mathematical functions (real, complex, and vector). Across all examples, our numerical experiments suggest that the new methods converge for a significantly wider range of initial guesses with minimal increase in computational cost. Given the ubiquity of Newton's method, an improvement in its applicability and convergence is a significant step forward, and will reduce computation times several-folds across many disciplines. Additionally, this multiplicative transformation may improve other techniques where a linear approximation is used.

Gareth Jones, Jim Parr, Perumal Nithiarasu et al.

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

Gareth Jones, Jim Parr, Perumal Nithiarasu et al.

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.

Damiano Lombardi, Sanjay Pant

A non-parametric k-nearest neighbour based entropy estimator is proposed. It improves on the classical Kozachenko-Leonenko estimator by considering non-uniform probability densities in the region of k-nearest neighbours around each sample point. It aims at improving the classical estimators in three situations: first, when the dimensionality of the random variable is large; second, when near-functional relationships leading to high correlation between components of the random variable are present; and third, when the marginal variances of random variable components vary significantly with respect to each other. Heuristics on the error of the proposed and classical estimators are presented. Finally, the proposed estimator is tested for a variety of distributions in successively increasing dimensions and in the presence of a near-functional relationship. Its performance is compared with a classical estimator and shown to be a significant improvement.