Portail des publications scientifiques IMT Mines Alès
Not a member yet
5198 research outputs found
Sort by
Relevance-Aware Thresholding in Online Conformal Prediction for Time Series
No full textInternational audienceUncertainty quantification has received considerable interestin recent works in Machine Learning. In particular, Conformal Prediction(CP) gains ground in this field. For the case of time series, OnlineConformal Prediction (OCP) becomes an option to address the problemof data distribution shift over time. Indeed, the idea of OCP is to updatea threshold of some quantity (whether the miscoverage level or the quantile)based on the distribution observation. To evaluate the performanceof OCP methods, two key aspects are typically considered: the coveragevalidity and the prediction interval width minimization. Recently,new OCP methods have emerged, offering long-run coverage guaranteesand producing more informative intervals. However, during the thresholdupdate step, most of these methods focus solely on the validity ofthe prediction intervals – that is, whether the ground truth falls insideor outside the interval – without accounting for their relevance. In thispaper, we aim to leverage this overlooked aspect. Specifically, we proposeenhancing the threshold update step by replacing the binary evaluation(inside/outside) with a broader class of functions that quantify the relevanceof the prediction interval using the ground truth. This approachhelps prevent abrupt threshold changes, potentially resulting in narrowerprediction intervals. Indeed, experimental results on real-world datasetssuggest that these functions can produce tighter intervals compared toexisting OCP methods while maintaining coverage validity
Enhancing WSI-Based Survival Analysis with Report-Auxiliary Self-distillation
No full textInternational audienceSurvival analysis based on Whole Slide Images (WSIs) is crucial for evaluating cancer prognosis, as they offer detailed microscopic information essential for predicting patient outcomes. However, traditional WSI-based survival analysis usually faces noisy features and limited data accessibility, hindering their ability to capture critical prognostic features effectively. Although pathology reports provide rich patient-specific information that could assist analysis, their potential to enhance WSI-based survival analysis remains largely unexplored. To this end, this paper proposes a novel Report-auxiliary self-distillation (Rasa) framework for WSI-based survival analysis. First, advanced large language models (LLMs) are utilized to extract fine-grained, WSI-relevant textual descriptions from original noisy pathology reports via a carefully designed task prompt. Next, a self-distillation-based pipeline is designed to filter out irrelevant or redundant WSI features for the student model under the guidance of the teacher model’s textual knowledge. Finally, a risk-aware mix-up strategy is incorporated during the training of the student model to enhance both the quantity and diversity of the training data. Extensive experiments carried out on our collected data (CRC) and public data (TCGA-BRCA) demonstrate the superior effectiveness of Rasa against state-of-the-art methods. Our code is available at https://github.com/zhengwang9/Rasa
Trajectories of interbrain synchrony during teamwork: links with team composition and performance
No full textInternational audienceTeamwork is a dynamic phenomenon that develops over time. Team cognition involved in teamwork is known to increase over successive task episodes. However, there is limited understanding of the trajectory of team cognition within a single task episode. The current functional near infrared spectroscopy hyperscanning study used interbrain synchrony (IBS) to investigate how team cognition develops throughout a task. The links of IBS trajectories with team personality composition and performance were also investigated. Results showed that (i) IBS did not significantly change over time within the sample; (ii) teams show significant differences in IBS trajectories; (iii) team personality did not predict IBS trajectories and; (iv) IBS trajectories did not predict team performance. While IBS was found to appear in our sample, these results warrant replication and additional research is required to better understand IBS trajectories, especially the heterogeneity across teams
Synthesis and functionalization of novel porous chitosan-bentonite hybrid for efficient removal of radioactive strontium – Improvement of sorption performance by sulfonation
No full textInternational audienceThe highly efficient 90Sr removal from contaminated effluents became a critical challenge for the development of nuclear activities. Herein, chitosan is associated with bentonite for manufacturing spherical composite beads (Ch-BEN). This material having weak affinity for strontium (hard acid according Pearson's principles), it was necessary functionalizing the support with hard base groups for enhancing metal removal. Ch-BEN was functionalized as S-Ch-BEN, by grafting 2-acrylamido-2-methyl-1-propanesulfonic acid (AMPS). The coexistence of N-bearing groups and sulfonic acid moieties (associated with tautomerization effects) was characterized by Fourier-Transform infrared and X-ray photoelectron spectroscopy for confirming chemical functionalization and for correlating sorption mechanisms. The sorption capacity (at the optimum pH; i.e., 7) was increased by 2.5 to 2.7-fold after AMPS grafting (up to 2.8 mmol Sr g−1). Sorption isotherms were fitted by the Sips equation, while uptake kinetics was improved and followed pseudo-first order rate equation. The sorption equilibrium was achieved in 60 min for Ch-BEN compared to 30 min for S-Ch-BEN. The functionalization (with sulfonic acid groups) not only improves maximum sorption capacity and kinetics but also selective strontium recovery (from equimolar multi-component solutions and seawater). The stability at sorbent recycling was considerably improved by AMPS grafting (achieved complete desorption with 0.3 M HNO3 solution; the loss at the 10th recycling <4 %). Spherical conditioning of the composite allowed testing the capacity of the sorbent for application in fixed-bed columns. The major issue concerning strontium deals with the removal of its radionuclide forms: a series of tests using 90Sr spiked solutions (of increasing complexity) confirmed the efficiency of S-Ch-BEN for recovering 90Sr, including for contaminated seawater. S-Ch-BEN reveals promising for environmental applications and treatment of low-level activity effluents
Towards Automating the Life Cycle Management of Digital Twins
No full textInternational audienceDigital Twins (DTs) are key to Industry 4.0 and IoT systems. By definition, as virtual counterparts of a Physical Object (PO), they have to be kept up-to-date with the PO at a fast pace, leading to multiple reconfigurations, reconnections and redeployments throughout their lifetime. On the one hand, there is a growing interest around DT platforms and DT modeling languages from both academia and industry. However, systematic operational methodologies and practical tools for managing DTs resources and life cycle are more scarce. This paper identifies the need for a domain-agnostic solution to systematically manage the DT life cycle. On the other hand, GitOps provides principles and concrete actionable tools, such as CI/CD pipelines, declarative cluster state and version control, to manage software architectures that needs to: (1) be redeployed swiftly without interruptions and, (2) be versioned. By combining the static Digital Twin Definition Language and the GitOps framework, a new way of automating life cycle management of DTs at runtime is proposed in this paper. A conceptualization of core DT life cycle operations at runtime such as connect, synchronize, or configure and a similar work on generic GitOps operations is provided. Then, a mapping from DT to generic GitOps operations is proposed. Finally, a proof-of-concept involving a Fischertechnik Training Factory 4.0 as a use-case is presented
Automatic Annotation of Drosophila Neuron Microscopy Images Using Generative AI
No full textInternational audienceThe extraction of metrics quantifying the morphology of the neuronal tree is essentialfor the study of neuronal development and function. However, the extraction of suchmorphometrics often requires manually tracing dendrites from high-resolution microscopyimages, a burdensome and time-consuming task restraining large-scale quantitative studies.In this work, we propose a method based on generative AI to automatically generateannotated microscopy images from existing open datasets of traced neurons. We showthat the annotated data can be used to train a deep-learning model to extract the neuronmorphology, mitigating the need for manual tracing. Finally, we show an application ofdeep-learning models to automatically extract morphometrics and analyze large cohorts ofsingle-neuron Drosophila microscopy images
Développement d'une méthode de mesure pour estimer la concentration en phase gazeuse à la surface des matériaux sources des COSV à n'importe quelle température
No full textCorrigendum : https://doi.org/10.1016/j.chemosphere.2025.144626International audienceSemi-volatile organic compounds (SVOCs) are ubiquitous toxic indoor pollutants. The gas-phase concentration in equilibrium with the material surface (y₀) is the key parameter to estimate their emissions from indoor materials. The Clausius-Clapeyron relationship was used to determine y₀ at 25 °C of phthalates (DnOP and DiNP), organophosphate (OP) esters (TCPP, TDCPP, TPP and EHDPP) and DINCH emitted by vinyl floorings (VF), polyvinyl chloride (PVC)-coated fabrics and polyurethane (PU) cushioning foam. These materials were placed in a Micro-Chamber/Thermal Extractor (M-CTE, MARKES) and the gas-phase concentrations were measured at several elevated temperatures (T). y₀ was estimated at each temperature and the empiric relationship between y₀ and T was established. The extrapolated y₀ values were compared to experimental values at 25 °C. The influence of the temperatures used and the position of the material in the cell were investigated. The results indicate that y₀ at 25 °C are very low, mostly below 1 μg m⁻³. While for DiNP, TCPP, TDCPP and DINCH, the results are of the same order of magnitude as those available in the literature, y₀ of DnOP, TPP and EHDPP are reported for the first time. The temperatures used had an impact on the extrapolated y₀, but the consequences should be limited due to low values. Upon lifting the materials in the cell, the steady state gas phase concentration was close to y₀. This configuration can be used for quick y₀ estimation. The benefit of this method is the ability to estimate y₀ at any temperature which is useful to estimate exposure under different scenarios.Les composés organiques semi-volatils (COSV) sont des polluants toxiques omniprésents dans les environnements intérieurs. La concentration en phase gazeuse en équilibre avec la surface du matériau (y₀) est un paramètre clé pour estimer leurs émissions à partir des matériaux présents en intérieur. La relation de Clausius-Clapeyron a été utilisée pour déterminer y₀ à 25 °C pour les phtalates (DnOP et DiNP), les esters organophosphorés (TCPP, TDCPP, TPP et EHDPP) et le DINCH émis par des revêtements de sol en vinyle (VF), des tissus enduits de PVC et des mousses de rembourrage en polyuréthane (PU). Ces matériaux ont été placés dans les micro-chambres (M-CTE, MARKES) et les concentrations en phase gazeuse ont été mesurées à plusieurs températures élevées (T). y₀ a été estimé à chaque température et une relation empirique entre y₀ et T a été établie. Les valeurs extrapolées de y₀ ont été comparées aux valeurs expérimentales à 25 °C. L’influence des températures utilisées et de la position du matériau dans la cellule a été étudiée. Les résultats indiquent que les valeurs de y₀ à 25 °C sont très faibles, généralement inférieures à 1 μg·m⁻³. Bien que pour le DiNP, le TCPP, le TDCPP et le DINCH, les résultats soient du même ordre de grandeur que ceux disponibles dans la littérature, les valeurs de y₀ pour le DnOP, le TPP et l’EHDPP sont rapportées pour la première fois. Les températures utilisées ont eu un impact sur les valeurs extrapolées de y₀, mais les conséquences devraient être limitées en raison des faibles valeurs. En surélevant les matériaux dans la cellule, la concentration en phase gazeuse à l’état stationnaire était proche de y₀. Cette configuration peut être utilisée pour une estimation rapide de y₀. L’avantage de cette méthode est la possibilité d’estimer y₀ à n’importe quelle température, ce qui est utile pour évaluer l’exposition dans différents scénarios
Entre thérapie manuelle et ostéopathie : enquête qualitative sur la spécialisation des kinésithérapeutes français
No full textInternational audienceIn the French healthcare system, physical therapy and osteopathy are the manual professions with the highest representation. In physical therapy, osteopathy, manual therapy and orthopedic manual therapy courses embody forms of ‘manual specificity’. This qualitative study, based on eight semi-structured interviews, explored the differences between these approaches, as well as the motivations of physiotherapists who specialize in them. It reveals issues of legitimacy, competition within the medical hierarchy, and a quest for professional identity. Despite their differences, these practices share common techniques and respond to an integrative trend in medicine
Personality change before and after diagnosis of chronic respiratory diseases.
No full textInternational audienceObjective: Chronic respiratory diseases can impact functioning and multiple physical and mental health outcomes, but their impact on psychological traits is less understood. This study investigated personality trajectories before and after the diagnosis of chronic respiratory diseases. Methods: Data were from the Health and Retirement Study. Personality and chronic respiratory conditions were assessed between 2006 and 2020 (N = 17,078; n = 1,044 with chronic respiratory diseases; 40,971 personality assessments). Multilevel models evaluated personality changes before and after chronic respiratory diseases diagnosis, accounting for sociodemographic, clinical, and behavioral factors and normative age-related trajectories.Results: Before the diagnosis of chronic respiratory diseases, small effects were detected for declines in extraversion, openness, agreeableness, and conscientiousness (effect sizes ranging from -0.10 for conscientiousness to -0.16 for openness, in standard deviations per decade). After the diagnosis of the disease, there was a steeper increase in neuroticism and steeper declines in extraversion, agreeableness, and conscientiousness compared to participants who never developed a chronic respiratory disease during the follow-up (effect sizes ranging from -0.29 for conscientiousness to 0.24 for neuroticism, in standard deviations per decade). Conclusions: Chronic respiratory diseases are associated with maladaptive personality changes, mainly following diagnosis. These findings highlight the importance of developing interventions to counter maladaptive personality trajectories in people with chronic respiratory conditions
Evolution of turbulent burning velocity and turbulence intensity during aluminum flame propagation
No full textInternational audienceBurning velocity, i.e. the consumption rate of the reactant by the flame front, is a key parameter for modelling flame propagation during accidental gas or dust explosions. Experimental investigations are thus needed to determine this burning velocity and to accurately validate numerical models. In this article, aluminum dust flame propagation inside a vertical tube is investigated. An innovative “direct method” is implemented to determine the burning velocity just ahead of the propagating flame front. This method is based on the Time-Resolved Particle Image Velocimetry (TR-PIV) technique to determine the unburned flow velocity. With this experimental setup, the turbulence intensity just ahead of the propagating flame front can thus also be estimated. Two granulometric distributions (with a median diameter of 6 μm and 20 μm), with three different dust concentrations for each granulometric distribution are investigated. An increase of the burning velocity with a decrease of the particle size is observed. Nevertheless, due to low levels of turbulence in front of the flame front during flame propagation, no significant influence of turbulence on burning velocity is observed during these experiments