All issues

The aim of study was to develop an ensemble machine learning method for constructing interpretable predictive models and to validate it using the example of predicting in-hospital mortality (IHM) in patients with ST-segment elevation myocardial infarction (STEMI).

A retrospective cohort study was conducted using data from 5446 electronic medical records of STEMI patients who underwent percutaneous coronary intervention (PCI). Patients were divided into two groups: 335 (6.2%) patients who died during hospitalization and 5111 (93.8%) patients with a favourable in-hospital outcome. A pool of potential predictors was formed using statistical methods. Through multimetric categorization (minimizing p-values, maximizing the area under the ROC curve (AUC), and SHAP value analysis), decision trees, and multivariable logistic regression (MLR), predictors were transformed into risk factors for IHM. Predictive models for IHM were developed using MLR, Random Forest Risk Factors (RandFRF), Stochastic Gradient Boosting (XGboost), Random Forest (RF), Adaptive boosting, Gradient Boosting, Light Gradient-Boosting Machine, Categorical Boosting (CatBoost), Explainable Boosting Machine and Stacking methods.

Authors developed the RandFRF method, which integrates the predictive outcomes of modified decision trees, identifies risk factors and ranks them based on their contribution to the risk of adverse outcomes. RandFRF enables the development of predictive models with high discriminative performance (AUC 0.908), comparable to models based on CatBoost and Stacking (AUC 0.904 and 0.908, respectively). In turn, risk factors provide clinicians with information on the patient’s risk group classification and the extent of their impact on the probability of IHM. The risk factors identified by RandFRF can serve not only as rationale for the prediction results but also as a basis for developing more accurate models.

A mathematical model of tumor growth in tissue taking into account angiogenesis and antiangiogenic therapy is developed. In the model the convective flows in tissue are considered as well as individual motility of tumor cells. It is considered that a cell starts to migrate if the nutrient concentration falls lower than the critical level and returns into proliferation in the region with high nutrient concentration. Malignant cells in the state of metabolic stress produce vascular endothelial growth factor (VEGF), stimulating tumor angiogenesis, which increases the nutrient supply. In this work an antiangiogenic drug which bounds irreversibly to VEGF, converting it to inactive form, is modeled. Numerical analysis of influence of antiangiogenic drug concentration and efficiency on tumor rate of growth and structure is performed. It is shown that antiangiogenic therapy can decrease the growth of low-invasive tumor, but is not able to stop it completely.

Modern mathematical models in the dynamics system “soil–plant” are considered. The components of this system are: agricultural plant, microorganisms of the rhizosphere (root zone of plants), the mineral nutrition elements of plants in their mobile and immobile forms. The model of submitted system based on the analysis of the adopted provisions was developed. The construction of system elements allows to display the coordinated dynamics of these elements among themselves. In particular, the dynamics of mineral nutrition elements in plants and the dynamics of their biomass are determined by the current contents in the rhizosphere of mineral fertilizers and organic origin substances (plant roots, leaves, etc.). The immobility of plants spatial distribution and the mobile spatial nature of microorganisms are assumed. This mechanism is determined by diffusion. Mutual relationships between weeds and pests are suggested. The dynamics of the mineral nutrition elements is determined by the peculiarity of sorption in the soil solution, environmental conditions, organic decomposition and fertilizer application. An analytical study for a system where each of the components is represented by only one species (fertilizer, the association of microorganisms and plants) was performed. An adaptation of the wave propagation model in the “resource–consumer” system (Kolmogorov–Petrovsky–Piskunov waves) has been developed for annual agricultural crops. The developed model has been adapted for the growth of Krasnoufimskaya-100 spring wheat in a vessel on peat lowland soil, where nitrogen, phosphorus, and potassium fertilizers were added variably. Sample distributions are plants biomass and the content of mineral nutrition elements in them. The parametric identification of the model and its adequacy was performed. An assessment of the model adequacy showed a good agreement between the model and experimental data.

Numerical modeling of shear-induced platelet activation in haemodialysis arteriovenous fistulas was carried out in this work. The goal was to investigate the mechanisms of threshold shear-induced platelet activation in fistulas. For shear-induced platelet activation to take place, shear stress accumulated by platelets along corresponding trajectories in blood flow had to exceed a definite threshold value. The threshold value of cumulative shear stress was supposed to depend on the multimer size of von Willebrand factor macromolecules acting as hydrodynamic sensors for platelets. The effect of arteriovenous fistulas parameters, such as the anastomotic angle, blood flow rate, and the multimer size of von Willebrand factor macromolecules, on platelet activation risk was studied. Parametric diagrams have been constructed that make it possible to distinguish the areas of parameters corresponding to the presence or absence of shear-induced platelet activation. Scaling relations that approximate critical curves on parametric diagrams were obtained. Analysis showed that threshold fistula flow rate is higher for obtuse anastomotic angle than for sharp ones. This means that a fistula with obtuse angle can be used in wider flow rate range without risk of platelet activation. In addition, a study of different anastomosis configurations of arteriovenous fistulas showed that the configuration “end of vein to end of artery” is among the safest. For all the investigated anastomosis configurations, the critical curves on the parametric diagrams were monotonically decreasing functions of von Willebrand factor multimer size. It was shown that fistula flow rate should have a significant impact on the probability of thrombus formation initiation, while the direction of flow through the distal artery did not affect platelet activation. The obtained results allowed to determine the safest fistula configurations with respect to thrombus formation triggering. The authors believe that the results of the work may be of interest to doctors performing surgical operations for creation of arteriovenous fistulas for haemodialysis. In the final section of the work, possible clinical applications of the obtained results by means of mathematical modeling are discussed.

This paper is dedicated to the analysis of medical and biological data obtained through locomotor training and testing of astronauts conducted both on Earth and during spaceflight. These experiments can be described as the astronaut’s movement on a treadmill according to a predefined regimen in various speed modes. During these modes, not only the speed is recorded but also a range of parameters, including heart rate, ground reaction force, and others, are collected. In order to analyze the dynamics of the astronaut’s condition over an extended period, it is necessary to perform a qualitative segmentation of their movement modes to independently assess the target metrics. This task becomes particularly relevant in the development of an autonomous life support system for astronauts that operates without direct supervision from Earth. The segmentation of target data is complicated by the presence of various anomalies, such as deviations from the predefined regimen, arbitrary and varying duration of mode transitions, hardware failures, and other factors. The paper includes a detailed review of several contemporary retrospective (offline) nonparametric methods for detecting multiple changepoints, which refer to sudden changes in the properties of the observed time series occurring at unknown moments. Special attention is given to algorithms and statistical measures that determine the homogeneity of the data and methods for detecting change points. The paper considers approaches based on dynamic programming and sliding window methods. The second part of the paper focuses on the numerical modeling of these methods using characteristic examples of experimental data, including both “simple” and “complex” speed profiles of movement. The analysis conducted allowed us to identify the preferred methods, which will be further evaluated on the complete dataset. Preference is given to methods that ensure the closeness of the markup to a reference one, potentially allow the detection of both boundaries of transient processes, as well as are robust relative to internal parameters.

The paper considers the problem of transverse impact on a thin preloaded fiber. The commonly accepted theory of transverse impact on a thin fiber is based on the classical works of Rakhmatulin and Smith. The simple relations obtained from the Rakhmatulin – Smith theory are widely used in engineering practice. However, there are numerous evidences that experimental results may differ significantly from estimations based on these relations. A brief overview of the factors that cause the differences is given in this article.

This paper focuses on the shear wave velocity, as it is the only feature that can be directly observed and measured using high-speed cameras or similar methods. The influence of the fiber preload on the wave speed is considered. This factor is important, since it inevitably arises in the experimental results. The reliable fastening and precise positioning of the fiber during the experiments requires its preload. This work shows that the preload significantly affects the shear wave velocity in the impacted fiber.

Numerical calculations were performed for Kevlar 29 and Spectra 1000 yarns. Shear wave velocities are obtained for different levels of initial tension. A direct comparison of numerical results and analytical estimations with experimental data is presented. The speed of the transverse wave in free and preloaded fibers differed by a factor of two for the setup parameters considered. This fact demonstrates that measurements based on high-speed imaging and analysis of the observed shear waves should take into account the preload of the fibers.

This paper proposes a formula for a quick estimation of the shear wave velocity in preloaded fibers. The formula is obtained from the basic relations of the Rakhmatulin – Smith theory under the assumption of a large initial deformation of the fiber. The formula can give significantly better results than the classical approximation, this fact is demonstrated using the data for preloaded Kevlar 29 and Spectra 1000. The paper also shows that direct numerical calculation has better corresponding with the experimental data than any of the considered analytical estimations.

A numerical simulation of fluid flow in a blood pump was performed using the FlowVision software package. This test problem, provided by the Center for Devices and Radiological Health of the US. Food and Drug Administration, involved considering fluid flow according to several design modes. At the same time for each case of calculation a certain value of liquid flow rate and rotor speed was set. Necessary data for calculations in the form of exact geometry, flow conditions and fluid characteristics were provided to all research participants, who used different software packages for modeling. Numerical simulations were performed in FlowVision for six calculation modes with the Newtonian fluid and standard $k-\varepsilon$ turbulence model, in addition, the fifth mode with the $k-\omega$ SST turbulence model and with the Caro rheological fluid model were performed. In the first stage of the numerical simulation, the convergence over the mesh was investigated, on the basis of which a final mesh with a number of cells of the order of 6 million was chosen. Due to the large number of cells, in order to accelerate the study, part of the calculations was performed on the Lomonosov-2 cluster. As a result of numerical simulation, we obtained and analyzed values of pressure difference between inlet and outlet of the pump, velocity between rotor blades and in the area of diffuser, and also, we carried out visualization of velocity distribution in certain cross-sections. For all design modes there was compared the pressure difference received numerically with the experimental data, and for the fifth calculation mode there was also compared with the experiment by speed distribution between rotor blades and in the area of diffuser. Data analysis has shown good correlation of calculation results in FlowVision with experimental results and numerical simulation in other software packages. The results obtained in FlowVision for solving the US FDA test suggest that FlowVision software package can be used for solving a wide range of hemodynamic problems.

The method for mapping of intermediate representations (IR) set of C, C++ programs to vector embedding space is considered to create an empirical estimation framework for static performance prediction using LLVM compiler infrastructure. The usage of embeddings makes programs easier to compare due to avoiding Control Flow Graphs (CFG) and Data Flow Graphs (DFG) direct comparison. This method is based on transformation series of the initial IR such as: instrumentation — injection of artificial instructions in an instrumentation compiler’s pass depending on load offset delta in the current instruction compared to the previous one, mapping of instrumented IR into multidimensional vector with IR2Vec and dimension reduction with t-SNE (t-distributed stochastic neighbor embedding) method. The D1 cache miss ratio measured with perf stat tool is considered as performance metric. A heuristic criterion of programs having more or less cache miss ratio is given. This criterion is based on embeddings of programs in 2D-space. The instrumentation compiler’s pass developed in this work is described: how it generates and injects artificial instructions into IR within the used memory model. The software pipeline that implements the performance estimation based on LLVM compiler infrastructure is given. Computational experiments are performed on synthetic tests which are the sets of programs with the same CFGs but with different sequences of offsets used when accessing the one-dimensional array of a given size. The correlation coefficient between performance metric and distance to the worst program’s embedding is measured and proved to be negative regardless of t-SNE initialization. This fact proves the heuristic criterion to be true. The process of such synthetic tests generation is also considered. Moreover, the variety of performance metric in programs set in such a test is proposed as a metric to be improved with exploration of more tests generators.

In the first part of the paper we present convergence analysis of AGMsDR method on a new class of functions — in general non-convex with $M$-Lipschitz-continuous gradients that satisfy Polyak – Lojasiewicz condition. Method does not need the value of $\mu^{PL}>0$ in the condition and converges linearly with a scale factor $\left(1 - \frac{\mu^{PL}}{M}\right)$. It was previously proved that method converges as $O\left(\frac1{k^2}\right)$ if a function is convex and has $M$-Lipschitz-continuous gradient and converges linearly with a~scale factor $\left(1 - \sqrt{\frac{\mu^{SC}}{M}}\right)$ if the value of strong convexity parameter $\mu^{SC}>0$ is known. The novelty is that one can save linear convergence if $\frac{\mu^{PL}}{\mu^{SC}}$ is not known, but without square root in the scale factor.

The second part presents modification of AGMsDR method for solving problems that allow alternating minimization (Alternating AGMsDR). The similar results are proved.

As the result, we present adaptive accelerated methods that converge as $O\left(\min\left\lbrace\frac{M}{k^2},\,\left(1-{\frac{\mu^{PL}}{M}}\right)^{(k-1)}\right\rbrace\right)$ on a class of convex functions with $M$-Lipschitz-continuous gradient that satisfy Polyak – Lojasiewicz condition. Algorithms do not need values of $M$ and $\mu^{PL}$. If Polyak – Lojasiewicz condition does not hold, the convergence is $O\left(\frac1{k^2}\right)$, but no tuning needed.

We also consider the adaptive catalyst envelope of non-accelerated gradient methods. The envelope allows acceleration up to $O\left(\frac1{k^2}\right)$. We present numerical comparison of non-accelerated adaptive gradient descent which is accelerated using adaptive catalyst envelope with AGMsDR, Alternating AGMsDR, APDAGD (Adaptive Primal-Dual Accelerated Gradient Descent) and Sinkhorn's algorithm on the problem dual to the optimal transport problem.

Conducted experiments show faster convergence of alternating AGMsDR in comparison with described catalyst approach and AGMsDR, despite the same asymptotic rate $O\left(\frac1{k^2}\right)$. Such behavior can be explained by linear convergence of AGMsDR method and was tested on quadratic functions. Alternating AGMsDR demonstrated better performance in comparison with AGMsDR.

Laminar-turbulent transition is the subject of an active research related to improvement of economic efficiency of air vehicles, because in the turbulent boundary layer drag increases, which leads to higher fuel consumption. One of the directions of such research is the search for efficient methods, that can be used to find the position of the transition in space. Using this information about laminar-turbulent transition location when designing an aircraft, engineers can predict its performance and profitability at the initial stages of the project. Traditionally, $e^N$ method is applied to find the coordinates of a laminar-turbulent transition. It is a well known approach in industry. However, despite its widespread use, this method has a number of significant drawbacks, since it relies on parallel flow assumption, which limits the scenarios for its application, and also requires computationally expensive calculations in a wide range of frequencies and wave numbers. Alternatively, flow analysis can be done by using Dynamic Mode Decomposition, which allows one to analyze flow disturbances using flow data directly. Since Dynamic Mode Decomposition is a dimensionality reduction method, the number of computations can be dramatically reduced. Furthermore, usage of Dynamic Mode Decomposition expands the applicability of the whole method, due to the absence of assumptions about the parallel flow in its derivation.

The presented study proposes an approach to finding the location of a laminar-turbulent transition using the Dynamic Mode Decomposition method. The essence of this approach is to divide the boundary layer region into sets of subregions, for each of which the transition point is independently calculated, using Dynamic Mode Decomposition for flow analysis, after which the results are averaged to produce the final result. This approach is validated by laminar-turbulent transition predictions of subsonic and supersonic flows over a 2D flat plate with zero pressure gradient. The results demonstrate the fundamental applicability and high accuracy of the described method in a wide range of conditions. The study focuses on comparison with the $e^N$ method and proves the advantages of the proposed approach. It is shown that usage of Dynamic Mode Decomposition leads to significantly faster execution due to less intensive computations, while the accuracy is comparable to the such of the solution obtained with the $e^N$ method. This indicates the prospects for using the described approach in a real world applications.