Paul Thiry's thesis defense

Summary

Nonspecific low back pain (NLBP) exhibits significant variability, both in its clinical presentation and in the lumbar movements of affected patients. This variability is inherent to human adaptability and may be crucial for the development of diagnostic and rehabilitation strategies. Analyzing the variability of lumbar movements offers new perspectives for the management of patients with LNS. Inertial sensors and non-linear analysis methods have the potential to provide a more accurate and personalized assessment of the condition. The overall aim of this thesis was to verify the relevance of integrating kinematic sensors and nonlinear motion analysis, in the biopsychosocial clinical examination of patients with LNS.

To meet this challenge, it is necessary to adopt a comprehensive approach that should imperatively take into account the biopsychosocial factors influencing LNS and integrate a thorough understanding of lumbar movement variability and its role in human adaptability for optimal LNS management.

This objective initially required, after the development of a low-cost kinematic data recording system, the choice of a method for analyzing the generated time series. Among the many non-linear tools available, entropy emerged as an essential indicator for quantifying the complexity of time series. In this context, entropy can be used to measure the uncertainty and complexity of biological systems, offering a new perspective on the analysis of clinical data. Sample entropy (SampEn), in particular, has emerged as a relevant complexity measure for short time series. By combining variability and complexity analysis, it should be possible to obtain a deeper understanding of movement patterns and motor control perturbations in the context of low back pain.

The first step was to analyze time series by developing a simple test, feasible in clinical practice and of sufficient duration to harvest time series for SampEn calculation, the "bend and return test" (b&r test). The reproducibility of this test was then verified by two studies in the second stage. A third stage verified that the b&r test was capable of differentiating a healthy population from one suffering from non-specific chronic low-back pain, with the help of artificial intelligence. Finally, after developing an application (the NOMADe App) capable of digitizing and recording clinical data from the anamnesis and physical examination of patients with NSL, and adding scores from various validated questionnaires, it was possible to compare these digitized data with variability and kinematic complexity data. These comparisons highlighted the possibility of very precisely phenotyping each patient within a biopsychosocial framework, and providing the therapist with precise information for individualized management. Machine Learning analysis of the kinematic data also provided a prediction of the results of validated questionnaires, enabling the therapist to be guided without loss of time towards the choice of questionnaires relevant to the patient. Our findings call for precision physiotherapy and suggest a shift to evidence-based personalized practice.

Keywords

inertial sensor, low back pain, sample entropy, variability, physiotherapy, artificial intelligence.

Jury composition

Pr DUMAS Raphaël, Université Gustave Eiffel, France (Rapporteur)
Dr GERUS Pauline, Université Côte d'Azur, France (Rapporteur)
Pr DEMOULIN Christophe, Université de Liège, Belgium (Examiner)
Pr ARMAND, Stéphane, University of Geneva, Switzerland (Examiner)
Pr SIMONEAU-BUSSINGER Emilie, Université Polytechnique Hauts-de-France (Thesis Director)
Pr THÉVENON André, MPR, Université Lille (Thesis co-director)
Pr TIFFREAU Vincent, MPR, Université Lille (Guest)