The research activities of the Applied Statistics Team (ESA) are driven by the greater need to model the behavior of complex processes from numeric data, be it in the industry or in the biological field. Thus, the Applied Statistics Team develops theoretical and practical tools suited to the processing of this type of data and to the type of complexity that has to be mastered, such as high nonlinearities, the large dimensionality of the phenomena, or the uncertainty surrounding them.

Since its creation by Pierre-Gilles de Gennes in March 2000, ESA has consolidated two large fields of operation, the development of statistical methods for engineering sciences on one hand, and the modeling of nervous systems on the other hand. In both fields, ESA has developed expertise in the implementation of artificial neural networks.

Since its incorporation into the Experimental and Clinical Respiratory Neurophysiology unit of the Pitié-Salpêtrière, UMRS1158, ESA lends its mathematical and numerical tools to the main research axes of this unit, e.g. (i) statistical tools, such as design of experiments, analyses and tests applied to experimental data, (ii) time-frequency analysis tools for physiological signals, and (iii) modeling and simulation tools, such as artificial neural networks.


Modeling of the living
Study of the links between activity and structure in biologically relevant neural network architectures. Biological modeling and simulation. Examples:
• Modeling of the genesis of ventilatory rhythms in vertebrates
• Modeling of the bio-mechanical properties of the upper airway in humans

Biological signal processing and analysis
Time-frequency analyses. Evaluation and construction of relevant descriptors and multivariate analyses. Examples:
• Analysis of cardiac signals for the study of the interactions between cardiac activity, autonomic nervous system and respiration
• Removal of the cardiac artefacts from diaphragmatic electromyograms
• Analysis of ventilatory signals for the dynamic characterization of chronic obstructive pulmonary disease (COPD) patients during exercise
• Analysis of ventilatory signals for the description in terms of apneas/hypopneas of neurograms obtained on a mutant mouse

Biostatistical anlyses
Multivariate analyses: generalized linear regressions, unsupervised methods. Examples:
• Study of the survival of patients with primary malignant brain tumors admitted to the intensive care unit with or without acute respiratory distress
• Characterization of the sensory and emotional perceptions of dyspnoeic patients using unsupervised clustering
• Analysis of the effects of stress in the rat concerning the cardiac measurements of the autonomic nervous system balance and the response to hypercapnia/hypoxia, with or without blocking of the orexin neurons
• Contribution to the elucidation of epigenetic mechanisms in cystic fibrosis using statistical analyses of the methylation of CFTR modifying genes, transcriptomic data, etc., in collaboration with the Genetics laboratory of rare diseases, UMRS 827 (Montpellier)
• Design and analysis of transcriptome experiments for the study of Down’s syndrome and Alzheimer’s disease, in collaboration with the Brain & Spine Institute
• Design and analysis of transcriptome experiments for the elucidation the mechanisms underlying neuropathic pain, in collaboration with the Brain Plasticity Laboratory of the ESPCI

Statistical methods for engineers
Development of construction and selection procedures for linear and nonlinear models (e.g. artificial neural networks). Bayesian statistics. Examples:
• Development of statistical tools for the analysis of GCxGC-MS data, in collaboration with the Laboratoire Sciences Analytiques, Bioanalytiques et Miniaturisation of the ESPCI, and the Criminal Research Institute of the French Gendarmerie, both for the processing of the chromatograms and for forensic applications such as the...

> More...


Journal papers

M. Decavèle, I. Rivals, C. Marois, M. Cantier, N. Weiss, L. Lemasle, H. Prodanovic, A. Idbaih,... (...)

Conferences and posters

I. Rivals, V. Cuzuel, G. Cognon, R. Leconte, D. Thiébaut, C. Sauleau & J. Vial
GC×GC-MS ... (...)


L. Personnaz & I. Rivals (2003)
Artificial neural networks for modeling, control and... (...)

> More publications...