Search

Among the different indicators that quantify the spread of an epidemic such as the on-going COVID-19, stands first the reproduction number which measures how many people can be contaminated by an infected person. In order to permit the monitoring of the evolution of this number, a new estimation procedure is proposed here, assuming a well-accepted model for current incidence data, based on past observations. The novelty of the proposed approach is twofold: 1) the estimation of the reproduction number is achieved by convex optimization within a proximal-based inverse problem formulation, with constraints aimed at promoting piecewise smoothness; 2) the approach is developed in a multivariate setting, allowing for the simultaneous handling of multiple time series attached to different geographical regions, together with a spatial (graph-based) regularization of their evolutions in time. The effectiveness of the approach is first supported by simulations, and two main applications to real COVID-19 data are then discussed. The first one refers to the comparative evolution of the reproduction number for a number of countries, while the second one focuses on French departments and their joint analysis, leading to dynamic maps revealing the temporal co-evolution of their reproduction numbers. [ABSTRACT FROM AUTHOR]

Objective: Numerous indices were devised for the statistical characterization of temporal dynamics of heart rate variability (HRV) with the aim to discriminate between healthy subjects and nonhealthy patients. Elaborating on the concepts of (multi)fractal and nonlinear analyses, the present contribution defines and studies formally novel non Gaussian multiscale representations. Methods: A methodological framework for non Gaussian multiscale representations constructed on wavelet p-leaders is developed, relying a priori neither on exact scale-free dynamics nor on predefined forms of departure from Gaussianity. Its versatility in quantifying the strength and nature of departure from Gaussian is analyzed theoretically and numerically. The ability of the representations to discriminate between healthy subjects and congestive heart failure (CHF) patients, and between survivors and nonsurvivor CHF patients, is assessed on a large cohort of 198 subjects. Results: The analysis leads to conclude that i) scale-free and multifractal dynamics are observed, both for healthy subjects and CHF patients, for time scales shorter than ${\text{170 s}}$ ; ii) a circadian evolution of multifractal and non Gaussian properties of HRV is evidenced for healthy subjects, but not for CHF patients; iii) non Gaussian multiscale indices possess high discriminative abilities between survivor and nonsurvivor CHF patients, at specific time scales ($\sim \!{\text{20 s}}$ and $\sim\! {\text{85 s}}$). Conclusions: The non Gaussian multiscale representations provide evidence for the existence of short-term cascade-type multifractal mechanisms underlying HRV for both healthy and CHF subjects. A circadian evolution of this mechanism is only evidenced for the healthy group, suggesting an alteration of the sympathetic-parasympathetic balance for CHF patients. Significance: Results obtained for a large cohort of subjects suggest that the novel non Gaussian indices might robustly quantify crucial information for clinical risk stratification in CHF patients. [ABSTRACT FROM AUTHOR]

Inspired from modern out-of-equilibrium statistical physics models, a matrix product based framework is defined and studied, that permits the formal definition of random vectors and time series whose desired joint distributions are a priori prescribed. Its key feature consists of preserving the writing of the joint distribution as the simple product structure it has under independence, while inputing controlled dependencies amongst components: This is obtained by replacing the product of probability densities by a product of matrices of probability densities. It is first shown that this matrix product model can be remapped onto the framework of Hidden Markov Models. Second, combining this dual perspective enables us both to study the statistical properties of this model in terms of marginal distributions and dependencies (a stationarity condition is notably devised) and to devise an efficient and accurate numerical synthesis procedure. A design procedure is also described that permits the tuning of model parameters to attain targeted statistical properties. Pedagogical well-chosen examples of times series and multivariate vectors aim at illustrating the power and versatility of the proposed approach and at showing how targeted statistical properties can actually be prescribed. [ABSTRACT FROM AUTHOR]

Piecewise constant denoising can be solved either by deterministic optimization approaches, based on the Potts model, or by stochastic Bayesian procedures. The former lead to low computational time but require the selection of a regularization parameter, whose value significantly impacts the achieved solution, and whose automated selection remains an involved and challenging problem. Conversely, fully Bayesian formalisms encapsulate the regularization parameter selection into hierarchical models, at the price of high computational costs. This contribution proposes an operational strategy that combines hierarchical Bayesian and Potts model formulations, with the double aim of automatically tuning the regularization parameter and maintaining computational efficiency. The proposed procedure relies on formally connecting a Bayesian framework to a $\ell _2$ -Potts functional. Behaviors and performance for the proposed piecewise constant denoising and regularization parameter tuning techniques are studied qualitatively and assessed quantitatively, and shown to compare favorably against those of a fully Bayesian hierarchical procedure, both in accuracy and computational load. [ABSTRACT FROM PUBLISHER]