Inverse modelling of an aneurysm’s stiffness using surrogate-based optimization and fluid-structure interaction simulations

Characterization of the mechanical properties of arterial tissues is highly relevant. In this work, we apply an inverse modelling approach to a model accounting for an aneurysm and the distal part of the circulation which can be modified using two independent stiffness parameters. For given values of these parameters, the position of the arterial wall as a function of time is calculated using a forward simulation which takes the fluid-structure interaction (FSI) into account. Using this forward simulation, the correct values of the stiffness parameters are obtained by minimizing a cost function, which is defined as the difference between the forward simulation and a measurement. The minimization is performed by means of surrogate-based optimization using a Kriging model combined with the expected improvement infill criterion. The results show that the stiffness parameters converge to the correct values, both for a zero-dimensional and for a three-dimensional model of the aneurysm.     

Primary high-dose intradermal hepatitis B vaccination in hemodialysis: cost-effectiveness evaluation at 2 years

Reinforced hepatitis B (HB) vaccination schedules have been tested in nonresponsive hemodialysis (HD) patients. Primary high-dose intradermal (ID) vaccination in HD has been proposed in one study with higher seroconversion rate, but no cost analysis was made. The aim of this prospective study was to confirm this previous report and focus on a cost-effectiveness evaluation of the thorough vaccination with a maintenance program. Thirty-five chronic incident HD patients received primary ID HB vaccination with a reinforced schedule (20 microg Engerix-B every 2 weeks). Revaccination with a monthly single ID dose of 20 microg was performed whenever anti-HBs titer fell under 20 IU/L and continued until a titer of 20 U/L was reached. Outcome measures were cumulative seroconversion rates, mean levels of anti-HBs, maintenance booster doses, rate of seroprotection at the end of the 2-year follow-up and subsequent costs. The present study was associated with an earlier peak of anti-HBs titer (3.9+/-1.7 months) and a higher cumulative seroconversion rate (96.9%) after 1 year. Moreover, a low-booster shot (17.4 microg) of ID Engerix-B/year/patient confers a 100% seroprotection for all responders for a second-year period. The mean cost of our schedule is 127.7 euro/patient for a 2-year period, revaccination included. This current study demonstrates that primary reinforced ID HB vaccination with a maintenance program for a 2-year period warrants the best cost-effectiveness ratio with rapid and sustained seroprotection in almost all HD patients.     

Identification of quasi-optimal regions in the design space using surrogate modeling

The use of Surrogate Based Optimization (SBO) is widely spread in engineering design to find optimal performance characteristics of expensive simulations (forward analysis: from input to optimal output). However, often the practitioner knows a priori the desired performance and is interested in finding the associated input parameters (reverse analysis: from desired output to input). A popular method to solve such reverse (inverse) problems is to minimize the error between the simulated performance and the desired goal. However, there might be multiple quasi-optimal solutions to the problem. In this paper, the authors propose a novel method to efficiently solve inverse problems and to sample Quasi-Optimal Regions (QORs) in the input (design) space more densely. The development of this technique, based on the probability of improvement criterion and kriging models, is driven by a real-life problem from bio-mechanics, i.e., determining the elasticity of the (rabbit) tympanic membrane, a membrane that converts acoustic sound wave into vibrations of the middle ear ossicular bones.     

Time series classification for the prediction of dialysis in critically ill patients using echo statenetworks

Objective: Time series often appear in medical databases, but only few machine learning methods exist that process this kind of data properly. Most modeling techniques have been designed with a static data model in mind and are not suitable for coping with the dynamic nature of time series. Recurrent neural networks (RNNs) are often used to process time series, but only a few training algorithms exist for RNNs which are complex and often yield poor results. Therefore, researchers often turn to traditional machine learning approaches, such as support vector machines (SVMs), which can easily be set up and trained and combine them with feature extraction (FE) and selection (FS) to process the high-dimensional temporal data. Recently, a new approach, called echo state networks (ESNs), has been developed to simplify the training process of RNNs. This approach allows modeling the dynamics of a system based on time series data in a straightforwardway.The objective of this study is to explore the advantages of using ESN instead of other traditional classifiers combined with FE and FS in classification problems in the intensive care unit (ICU) when the input data consists of time series. While ESNs have mostly been used to predict the future course of a time series, we use the ESN model for classification instead. Although time series often appear in medical data, little medical applications of ESNs have been studiedyet.Methods and material: ESN is used to predict the need for dialysis between the fifth and tenth day after admission in the ICU. The input time series consist of measured diuresis and creatinine values during the first 3days after admission. Data about 830 patients was used for the study, of which 82 needed dialysis between the fifth and tenth day after admission. ESN is compared to traditional classifiers, a sophisticated and a simple one, namely support vector machines and the naive Bayes (NB) classifier. Prior to the use of the SVM and NB classifier, FE and FS is required to reduce the number of input features and thus alleviate the curse dimensionality. Extensive feature extraction was applied to capture both the overall properties of the time series and the correlation between the different measurements in the time series. The feature selection method consists of a greedy hybrid filter-wrapper method using a NB classifier, which selects in each iteration the feature that improves prediction the best and shows little multicollinearity with the already selected set. Least squares regression with noise was used to train the linear readout function of the ESN to mitigate sensitivity to noise and overfitting. Fisher labeling was used to deal with the unbalanced data set. Parameter sweeps were performed to determine the optimal parameter values for the different classifiers. The area under the curve (AUC) and maximum balanced accuracy are used as performance measures. The required execution time was also measured.Results: The classification performance of the ESN shows significant difference at the 5% level compared to the performance of the SVM or the NB classifier combined with FE and FS. The NB+FE+FS, with an average AUC of 0.874, has the best classification performance. This classifier is followed by the ESN, which has an average AUC of 0.849. The SVM+FE+FS has the worst performance with an average AUC of 0.838. The computation time needed to pre-process the data and to train and test the classifier is significantly less for the ESN compared to the SVM andNB.Conclusion: It can be concluded that the use of ESN has an added value in predicting the need for dialysis through the analysis of time series data. The ESN requires significantly less processing time, needs no domain knowledge, is easy to implement, and can be configured using rules ofthumb.     

Broadband macromodelling of passive components using orthonormal vector fitting

Vector fitting is widely accepted as a robust macromodelling tool for efficient frequency domain identification of passive components. The orthonormal vector fitting technique is introduced, which improves the numerical stability of the method, by using orthonormal rational functions. This leads to better conditioned equations, reduces the numerical sensitivity to the choice of starting poles significantly, limits the number of required iterations, and reduces the overall computation time.     

A comparative study of kriging variants for the optimization of a turbomachinery system

Kriging is a well-established approximation technique for deterministic computer experiments. There are several Kriging variants and a comparative study is warranted to evaluate the different performance characteristics of the Kriging models in the computational fluid dynamics area, specifically in turbomachinery design where the most complex flow situations can be observed. Sufficiently accurate flow simulations can take a long time to converge. Hence, this type of simulation can benefit hugely from the computational cheap Kriging models to reduce the computational burden. The Kriging variants such as ordinary Kriging, universal Kriging and blind Kriging along with the commonly used response surface approximation (RSA) model were used to optimize the performance of a centrifugal impeller using CFD analysis. A Reynolds-averaged Navier–Stokes equation solver was utilized to compute the objective function responses. The responses along with the design variables were used to construct the Kriging variants and RSA functions. A hybrid genetic algorithm was used to find the optimal point in the design space. It was found that the best optimal design was produced by blind Kriging, while the RSA identified the worst optimal design. By changing the shape of the impeller, a reduction in inlet recirculation was observed, which resulted into an increase in efficiency.     

Parameterized S-Parameter Based Macromodeling With Guaranteed Passivity

This letter presents a novel parametric macromodeling technique for scattering input-output representations parameterized by design variables such as geometrical layout or substrate features. It provides accurate multivariate macromodels that are stable and passive by construction over the entire design space. Overall stability and passivity of the parametric macromodel are guaranteed by an efficient and reliable combination of rational identification and interpolation schemes based on a class of positive interpolation operators.     

Parametric macromodeling based on amplitude and frequency scaled systems with guaranteed passivity

We propose a novel parametric macromodeling method for systems described by scattering parameters, which depend on multiple design variables such as geometrical layout or substrate features. It is able to build accurate multivariate macromodels that are stable and passive over the entire design space. Poles and residues are parameterized indirectly. The proposed method is based on an efficient and reliable combination of rational identification, a procedure to find amplitude and frequency scaling system coefficients and positive interpolation schemes. Pertinent numerical examples validate the proposed parametric macromodeling technique. Copyright © 2011 John Wiley & Sons, Ltd.     

Accurate macromodeling algorithm for time domain identification of transient port responses

Orthonormal vector fitting is a robust method for broadband macromodeling of frequency domain responses. The use of orthonormal rational basis functions makes the conditioning of the system equations less sensitive to the initial pole specification when compared with the classical Vector Fitting procedure. This paper presents a time domain generalization of the technique to compute broadband rational macromodels from transient input–output port responses. The efficacy of the approach is illustrated by two numerical examples. Copyright © 2011 John Wiley & Sons, Ltd.     

Combining Krylov subspace methods and identification-based methods for model order reduction

Many different techniques to reduce the dimensions of a model have been proposed in the near past. Krylov subspace methods are relatively cheap, but generate non-optimal models. In this paper a combination of Krylov subspace methods and orthonormal vector fitting (OVF) is proposed. In that way a compact model for a large model can be generated. In the first step, a Krylov subspace method reduces the large model to a model of medium size, then a compact model is derived with OVF as a second step. Copyright © 2007 John Wiley & Sons, Ltd.