Portail HAL UPEC
Not a member yet
55480 research outputs found
Sort by
Vascular Geometry Drives Stroke Risk in Sickle Cell Disease
No full textInternational audienceSickle cell disease (SCD) is the leading cause of stroke in children and young adults, primarily due to cerebral vasculopathy (CV) occurring within the first decade of life. The main risk factor for CV is elevated blood velocity in intracranial arteries, contributing to stenosis formation in very young children. This study addresses three key questions: (i) the relationship between hemoglobin levels and intracranial blood velocities in SCD patients, (ii) additional factors contributing to elevated velocity beyond anemia, and (iii) the presence of flow anomalies. To investigate these aspects, biological and transcranial Doppler data from pediatric and adult SCD patients were analyzed. An image-based in silico modeling approach was also developed to simulate blood flow in the internal carotid, anterior cerebral, and middle cerebral arteries of SCD patients, of different age classes, and prior to any possible stenosis. Analysis revealed that while anemia is a recognized CV risk factor, it does not fully explain elevated velocities, as no significant correlation was found in children under five. In in silico simulations, young patients reached pathological arterial intracranial velocities at physiological flow rates, whereas adults remained below risk thresholds even at high flow rates. Pathological velocities were primarily observed in distal internal carotid arteries, where stenoses often develop. High flow rates, small arterial diameters, and pronounced curvatures led to extreme velocities and complex flow, likely causing endothelial damage and promoting CV progression. These findings enhance understanding of hemodynamic mechanisms underlying SCD-related stroke risk, paving the way for improved predictive models and early interventions
Relire les stratégies de formation à partir des couples de postures :: Analyse comparative de deux dispositifs en mathématiques avec les technologies numériques
No full textIn this paper, we revisit teacher-education strategies in mathematics through an analytical unit that articulates the trainer’s epistemological stance and the teachers’ professional stances during training. Building on Houdement and Kuzniak’s (1996) typology of training strategies, we conduct a comparative analysis of two short primary-level in-service programmes (Trainings A and B) based on chronological coding into episodes. Each episode is characterised by a stance pair crossing the trainer’s dominant epistemological stance (technician, didactician, epistemologist, facilitator) with the teachers’ dominant professional stance (pupil, student, teacher, teacher-collaborator). A configurational analysis using a 4×4 FE×PE matrix, together with a temporal analysis of the orchestration of stance pairs, highlights a homologous strategy in Houdement & Kuzniak’s (1996) sense, with a technician-dominant orientation in both programmes, but with distinct internal forms: relative dispersion and micro-shifts toward the teacher stance in Training A, and a strong polarisation around an immersive technician-pupil configuration in Training B. We thus hypothesise that training strategies, in these cases, should not be treated as homogeneous entities; rather, they are enacted as stabilised configurations of stance pairs distributed and orchestrated over time. The paper proposes a methodological formalisation that complements Houdement and Kuzniak’s (1996) typology and opens a comparative perspective for the study of more complex, especially collaborative, professional-development designs.Dans cet article, nous proposons une relecture des stratégies de formation des enseignants en mathématiques à partir d'une unité d'analyse articulant postures épistémologiques du formateur et postures professionnelles des enseignants en formation. En nous appuyant sur la typologie des stratégies de formation proposée par Houdement et Kuzniak (1996), nous analysons comparativement deux dispositifs courts du premier degré (formations A et B) à partir d'un codage chronologique en épisodes. Chaque épisode est caractérisé par un couple de postures croisant la posture épistémologique dominante du formateur (technicien, didacticien, épistémologue, accompagnateur) et la posture professionnelle dominante des enseignants (élève, étudiant, enseignant, enseignant-collaborateur). L'analyse configurationnelle, à partir d'une matrice FE×PE (4×4), et l'analyse temporelle, à partir de l'orchestration des couples, mettent en évidence une homologie, au sens de Houdement & Kuzniak (1996), à dominante technicienne dans les deux dispositifs, mais selon des formes internes distinctes : dispersion relative et micro-déplacements vers la posture d'enseignant dans la formation A, polarisation massive sur une immersion technicienne en posture d'élève dans la formation B. Nous formulons ainsi l'hypothèse que les stratégies de formation, dans ces deux cas, ne se donnent pas comme des entités homogènes mais se réalisent concrètement comme des configurations stabilisées de couples de postures distribuées et orchestrées dans le temps. L'article propose une formalisation méthodologique complémentaire à la typologie de Houdement et Kuzniak (1996) et ouvre une perspective comparative pour l'étude de dispositifs plus complexes, notamment collaboratifs.</div
Retours sur le 25e congrès national de la Société Française d’Étude et de Traitement de la Douleur
No full textInternational audienceABSTRACT Background Pain, whether as a symptom or a chronic disease, is the leading cause of medical consultations but there is no data on whether this is also the case in the practice of physiotherapy. Objective To assess the prevalence of patients with pain consulting a physiotherapist in France. Methods A questionnaire to be completed online was sent to physiotherapists practicing in France, asking them about the presence of pain in the patients they treated in the previous five days. The questionnaire was sent through representative healthcare organisations and professional societies, from January to September 2024. Results A total of 845 questionnaires were completed by physiotherapists, representing 52,497 patient consultations. Overall, 58.7% of patients seen by physiotherapists presented with pain, and pain was the main reason for consultation in 39.5% of patients. Physiotherapists reported a higher proportion of patients with chronic pain than with acute pain. Differences in pain characteristics were observed according to professional practice context: self‐employed physiotherapists more frequently reported pain located in the head or trunk and mixed acute and chronic pain profiles, whereas salaried physiotherapists more often reported acute pain and lower limb pain. Conclusion Pain is the major reason for physiotherapy consultations, and chronic pain is commonly encountered. Differences observed between self‐employed and salaried physiotherapists likely reflect variations in care contexts, referral pathways, and predominant fields of clinical activity. These results should inform health authorities of the role of physiotherapists in pain management and promote the development of standardised undergraduate and postgraduate pain education programs. Significance Statement This is the first nationwide and representative study to assess the prevalence and characteristics of pain among patients consulting physiotherapists. The results demonstrate that pain, particularly chronic pain, is a major reason for physiotherapy consultations in France. The evaluation of current educational programs and the implementation of standardised pain education for physiotherapists should now be undertaken to ensure best practices and ultimately improve patient outcomes
Long-term air pollution exposure and mental health in French adults of the CONSTANCES cohort: Role of black carbon independently of PM2.5
No full textInternational audienceAmbient air pollution could be associated with poor mental health. Black carbon (BC) has been highlighted as a crucial component of particulate matter; however, its isolated role independent from the total particulate matter mass has been poorly studied. Our study aimed to examine the associations between long-term exposure to particulate matter with a diameter <2.5 μm (PM2.5), BC and nitrogen dioxide (NO2) and psychological distress in the French CONSTANCES cohort and to assess the role of BC independently of PM2.5. This cross-sectional study included 104,146 adults. Psychological distress was assessed in 2019 using the General Health Questionnaire-12 (GHQ-12). Annual concentrations of PM2.5, BC, and NO2 estimated from land-use regression models at each participant's residential address. Negative binomial models with different covariate adjustments were used. A residuals method was used to assess the independent role of BC. Incident rate ratios (IRR) per an interquartile range (IQR) increase in exposure to each pollutant were calculated. Stratified analyses by age, sex, education and season were also conducted. Mean exposures were 9.38 μg/m3 for PM2.5 (IQR = 2.6), 1.15 × 10−5/m for BC (IQR = 0.5) and 19.1 μg/m3 for NO2 (IQR = 11.5). Exposure to each pollutant was significantly associated with higher psychological distress (IRR (95 % CI): 1.052 (1.014–1.092) for PM2.5, 1.078 (1.055–1.101) for BC, and 1.082(1.057–1.109) for NO2). Stronger associations were found for men, elderly, lower-educated, and during warm season. BC residuals were significantly associated with higher psychological distress when regressed on PM2.5 (1.055 (1.039–1.071)) and when regressed on NO2 (1.067(1.041–1.093)). Exposure to ambient air pollution was associated with psychological distress, with BC showing a deleterious role independently of PM2.5 and NO2
Tensor Decomposition-Driven Variational Autoencoder: Biomarker-Aware OCT Classification
No full textInternational audienceTensor decomposition techniques operate within Variational Autoencoders (VAEs) by restructuring the pa-rameterized components (e.g., weight matrices/tensors) and latent space into factorized, low-dimensionalrepresentations. Incorporating tensor decomposition techniques into VAEs allow for more efficient and in-terpretable modeling by addressing challenges such as parameter redundancy, computational inefficiency,and lack of structure in latent representations. We present a novel framework that enhances VAEs by in-corporating tensor decomposition techniques to boost both efficiency and performance. Building on the useof tensor decomposition in traditional autoencoders, we seamlessly integrate matrix and tensor factorizationmethods–specifically, nonnegative matrix factorization (NMF), Tucker decomposition (TD), and Neural net-work canonical polyadic decomposition (CPD-NN)–into the VAE architecture. Our primary model, symmetricVAE-CPDNN (sVAE-CPD-NN), achieves a significant 50% reduction in reconstruction error in the OCT Clas-sification dataset while decreasing parameter counts. On more complex datasets, it achieves up to a 30%reduction in parameters with only a marginal increase in error. Furthermore, our matrix-based variant (mVAE-CPD-NN) delivers competitive classification accuracy, underscoring the potential of tensor decompositiontechniques to optimize neural network design within probabilistic frameworks. To ease reproducibility, wemade the code available at this Github URL
Empty Dialogues? Problematizing Stakeholder Engagement under France’s Duty of Vigilance Law
No full textInternational audienc
Dual-Mode Functionalization of CeO2 Nanoparticles with Thiolated Dextran Sulfate for Enhanced Colloidal Stability and Redox Responsiveness
No full textInternational audienc
Long-time asymptotics for multivariate Hawkes processes with long-range interactions
No full textWe study a multivariate Hawkes process with long-range interactions, where the interaction strength decays as a power-law in the distance of the particles with exponent . Our main focus is on the long-time asymptotic behavior of the system. The proofs of our results combine techniques developed for short-range interactions, properties of -stable laws, and a Tauberian theorem. This model is more intricate and realistic for some applications, such as neural networks, where long-range connections are presen
Minimal tori in R^4
No full textWe describe tools for the study of minimal surfaces in R^4; some are classical (the Gauss maps) and some are newer (the link/braid/writhe at infinity). Then we look for complete proper non holomorphic minimal tori with total curvature −8π and a single end immersed in R^4. We translate the problem into a system of 10 quadratic or linear equations in 11 real variables with coefficients in terms of the Weierstrass function ℘ and give explicit solutions for these equations if T is a rectangular torus. For the square torus, we have a complete answer with a unique family of solutions generalizing the Chen-Gackstetter torus in R3. On the other hand, we show that there is no solution on the equianharmonic torus.30 pages, 4 figure