HAL Portal Univ-Côtedazur
Not a member yet
95391 research outputs found
Sort by
Meeting Healthcare Security Standards: From Legal Requirements to Technical Implementation
No full textInternational audienceEnsuring the security of healthcare data continues to pose a significant challenge, particularly for professionals operating beyond the confines of institutional frameworks. Existing tools frequently prove inadequate in meeting the regulatory requirements and usability expectations of these practitioners. This paper undertakes a comprehensive analysis of the legal and normative framework governing the protection of sensitive medical information. The analysis culminates in the derivation of a set of essential criteria that any compliant and practical solution must fulfill. We then proceeded to evaluate several secure messaging and health-focused platforms, highlighting their limitations with respect to these criteria. To address these criteria, we introduce U-HealthSec, an architecture is designed to ensure strong data protection while remaining accessible to non-expert users. U-HealthSec integrates established cryptographic methodologies and security protocols with regulatory compliance, thereby addressing the operational and legal requirements of healthcare professionals in a comprehensive and scalable manner
Embedded Experimental Platform Design for Multimodal Self-Mixing Interferometry
No full textstage de 5e année encadré par Laurent Rodriguez et Stéphane BarlandThis report presents the evolution of a self-mixing interferometry measurement system designed to estimate the speed of a target using embedded neural networks. Last year, work focused on migrating from an STM32 microcontroller to the RedPitaya platform, which offers better acquisition and processing capabilities. This year, the focus was on optimizing the RedPitaya system to stabilize real-time inference latency, as well as designing a multimodal architecture combining a truncated CNN and a self-organizing map (SOM) for dual-channel analysis. The models trained in this way can be easily deployed using a dedicated framework and application. These advances offer significant potential for applications in the biomedical field, where real-time processing and accuracy are essential.Ce rapport présente l’évolution d’un système de mesure par interférométrie auto-mélange destiné à estimer la vitesse d’une cible à l’aide de réseaux de neuronesembarqués. L’an dernier, les travaux ont consisté à migrer d’un microcontrôleur STM32 vers la plateforme RedPitaya, offrant de meilleures capacités d’acquisition et de traitement. Cette année, l’accent a été mis sur l’optimisation du système RedPitaya pour stabiliser la latence d’inférence en temps réel, ainsi que sur la conception d’une architecture multimodale combinant un CNN tronqué et une carte auto-organisatrice (SOM) pour l’analyse double canal. Les modèles ainsi entraînés peuvent être déployés facilement grâce à un framework et une application dédiée. Ces avancées offrent un potentiel important pour des applications dans le domaine biomédicale, où le traitement en temps réel et la précision sont essentiels
Deep learning methodolody for designing large scale matelenses in reflection
No full textInternational audienceDesigning efficient and portable virtual/augmented reality glasses is a goal that is being pursued by large companies envisioning the future of human-machine interfaces. The clearest path to miniaturize such devices is to evolve the optical system from refractive to diffractive, such as leveraging metalenses as a replacement for regular lenses. However, current methods for designing large scale metalenses demonstrates low efficiency for certain scenarios, for instance, high numerical aperture and off-axis illumination. In this study, we introduce a novel deep learning inverse design method utilizing our previously developed approach as a basis, incorporating substantial technical innovations to be applied to the design of large scale off-axis metalenses in reflection. Our method is based on the optimization of individual meta-deflectors and assembling them to form a metalens, and in this framework, a crucial requirement for obtaining high performance is assuring similarity among neighboring elements. To address this issue, we introduce a novel pre-trained tandem neural network with hard constraints. In total, we exploit 3 different neural networks. Firstly, a deep surrogate model for fast inference of the response of meta-deflectors and serve as a differentiable solver. Secondly, a Multi-Valued Artificial Neural Network to suggest geometrical parameter values given a target response. And finally, a tandem neural network with careful weight initialization to obtain the final design. Once the first two models are trained, the design of high-performance metalenses necessitates negligible computational effort. Our approach combines physical intuition about gratings with a state-of-the-art optimization methodology using modest computational effort to obtain highly efficient metalenses
Jean-François THÉMINES (2024), Les Écoles, leurs professeurs et le territoire. Guide pour une approche monographique , Lyon, Chronique sociale (155 p., 14 euros)
No full textDévelopper des indicateurs de comptabilité environnementale pour gérer la santé des cultures et des agroécosystèmes
No full textNational audienc
MetaboT: A Conversational AI-Agent for Accessible Mass Spectrometry Metabolomics Data Mining
No full textInternational audienceMass spectrometry metabolomics generates vast data requiring advanced methods for interpretation. Knowledge graphs (KGs) address these challenges by structuring mass spectrometry data, metabolite information, and their relationships into a network [PMID:38559298]. However, effective use of KGs demands detailed semantic modeling. To overcome this, we designed MetaboT, an AI system utilizing large language models (LLMs) to translate user questions into SPARQL, the query language of the KG. We demonstrated its effectiveness using the public KG endpoint for the largest open metabolomics dataset for plants [PMID:36649739].MetaboT employs specialized AI agents for handling user queries and interacting with the KG. The multi-agent system is constructed using the LangChain and LangGraph libraries, which facilitates the integration of LLMs with external tools. Its structured query generation process begins with specialized agents extracting necessary details such as entities and chemical taxonomies from user inputs. Next, a dedicated agent, informed of the KG ontology, crafts SPARQL query using these identifiers. The system then executes the query against the KG and returns structured, context-aware results.We assessed MetaboT’s performance on 50 metabolomics-related questions with expected answers. In addition to submitting these questions to MetaboT, we evaluated a baseline by submitting them to GPT-4o with a prompt that incorporated the KG ontology but did not provide specific entity IDs. This baseline achieved only 8% accuracy, compared to MetaboT’s 84%, underscoring the necessity of our multi-agent system for accurately retrieving entities and generating correct SPARQL queries.MetaboT shows promising performance, functioning as a conversational question-answering assistant. By enabling researchers to retrieve structured metabolomics data using natural language, MetaboT can facilitate data-driven discoveries. It automates SPARQL query generation and execution, which removes technical barriers that have traditionally limited access to KGs. MetaboT can be accessed on [https://metabot.holobiomicslab.eu/] and the code repository is available at [https://github.com/HolobiomicsLab/MetaboT]
Noninvasive ventilation in postoperative critically ill patients with morbid obesity: secondary analysis of the EXTUBOBESE multicentre randomised clinical trial
No full textInternational audienc
un consortium interuniversitaire réuni pour la reconnaissance par validation d'expérience (VAE) des savoirs issus de la vie avec la maladie/Situation de handicap
No full textNational audienceThe place of people living with at least one chronic illness and/or disability, as well as their numbers, have changed significantly in recent decades with advances in medical science and behavioral changes resulting from the proliferation of toxic environments, technologies, and lifestyles. Thus, over the last decade, patient knowledge (Jouet et al, 2010) both internationally and in France has become a social fact (Flora, 2012; Jouet, 2014) according to the definition of Emile Durkheim (1894), the father of French sociology. This social fact has also been recognized and highlighted by a renowned evidence-based medical journal, the British Medical Journal, which in 2013 published an editorial entitled: Let the patient revolution begin (Richards et al, 2013). This social event was also noted on a socioeconomic level in a health management science journal, which identified it as the blockbuster of the 21st century (Dentzer, 2013). In France, the laws on health democracy of March 4, 2002, HPST of August 9, 2009, mainly through Article 84 on the generalization of therapeutic patient education (TPE), on health democracy of January 26, 2016, and the improvement and transformation of the healthcare system of January 24, 2019, requiring medical schools to promote patient involvement in theoretical and practical training, have all encouraged the mobilization of users and patients to become active participants in both their own care and the healthcare system.These legislative contributions accompany changes within the healthcare system, where we can see that there has been a significant increase in social demand for patients, sick people, and people with disabilities to share their knowledge. As evidence of this, with regard to user representatives (RU) legitimized by the 2002 law more than 20 years ago, there are a number of unfilled positions and difficulties for medical schools and affiliated paramedical schools in organizing themselves to meet the demand set out in the 2019 law. A social mandate that led the Ministry of Health and Prevention and the Ministry of Higher Education and Research to create a steering committee (COPIL) in 2023 to produce recommendations in the form of operational procedures to support this legislative framework (Thomas et al, 2023). In 2024, COPIL gave way to the creation of a monitoring committee with the same intention. COPIL's first communication event took place during the 3rd legal and ethical research symposium on the place of the patient in the 21st century, in the era of patient care partnerships, as part of the inter-university research initiative launched in 2020 (2021; 2023) between the Legal Research Center of Paris 8 University (CRJP8) and the Center for Innovation in Partnership with Patients and the Public (CI3P) of the Faculty of Medicine of Côte d'Azur University.It is therefore in order to respond to the aforementioned social demand, which continues to grow at an inflationary rate, that, in parallel with the research initiative presented, an inter-university consortium has been formed in association with an association of “expert patients” in accordance with the social responsibility of universities.La place des personnes qui vivent avec au moins une maladie chronique et/ou une situation de handicap comme leur nombre se sont largement modifiés ces dernières décennies au fil des avancées des applications de la science en médecine, des transformations des comportements tant issus de la multiplication des environnements toxiques que des technologies et des modes de vie. Ainsi, au cours de la dernière décennie, le savoir des malades (Jouet et al, 2010) à l’international comme en France est devenu un fait social (Flora, 2012 ; Jouet ; 2014 ) selon la définition d’Emile Durkheïm (1894) , le père de la sociologie française. Ce fait social a d’ailleurs été pris en compte et souligné par une revue médicale réputée de médecine basée sur les preuves (Evidence Based Medicine), le British Medical Journal qui en 2013 publiait un édito sous ce titre : laissez la révolution des patients commencer (Richards et al, 2013) . Cet événement de société a également été remarqué sur un plan socioéconomique à travers une revue des sciences de gestion en santé dont un article l’identifie comme le blockbuster du 21ème siècle (Dentzer, 2013) . En France, les lois sur la démocratie sanitaire du 4 mars 2002, HPST du 9 août 2009 principalement à travers l’article 84 portant sur la généralisation de l’éducation thérapeutique du patient (ETP), de démocratie en santé du 26 janvier 2016, ou encore d’amélioration et de transformation du système de santé du 24 janvier 2019 demandant au UFRs de médecine de favoriser l’implication des patients dans les formations théoriques et pratiques, ont été autant de lois encourageant la mobilisation des usages comme des patients à devenir des acteurs tant de leurs soins qu’au sein du système de santé.Ces apports législatifs accompagnent des mutations au sein du système de santé dont nous pouvons constater qu’une commande sociale vis-à-vis de patients, malades et personnes en situation de handicap socialisant leurs savoirs s’est largement augmentée. Pour preuve vis-à-vis des représentants des usagers (RU) légitimés par la Loi de 2002, il y a plus de 20 ans, le nombre de postes non pourvus ou encore la difficulté des facultés de médecine et écoles paramédicales affiliées à s’organiser pour répondre à la demande cadrée en 2019 par la loi. Une commande sociale qui a amené en 2023 le ministère de la santé et de la prévention, et le ministère de l’enseignement supérieur et de la recherche à créer un COmité de PILotage (COPIL) afin de produire des recommandations sous forme de modalités opérationnelles pour accompagner ce cadre législatif (Thomas et al, 2023) . Un COPIL qui a laissé la place en 2024 à la création d’un comité de suivi toujours avec la même intention. Un COPIL dont la première manifestation de communication a été réalisée lors du 3ème colloque de recherche juridique et éthique sur la place du patient au 21ème siècle, à l’ère du partenariat de soin avec les patients dans le cadre de l’axe de recherche interuniversitaire initié en 2020 (2021 ; 2023 ) entre le Centre de Recherche Juridique de l’Université Paris 8 (CRJP8) et le Centre d’Innovation du Partenariat avec les Patients et le Public (CI3P) de la faculté de médecine d’Université Côte d’Azur.C’est donc pour faire face à la commande sociale évoquée qui continue de se développer selon une dynamique inflationniste, que parallèlement à l’axe de recherche présenté, s’est constitué un consortium interuniversitaire associé à une association de « patients experts » en respect de la responsabilité sociale des universités
Mazan. Anthropologie d’un procès pour viols
No full textInternational audienceSeptember 2024. The Mazan trial sent shockwaves through society. The judicial event spilled beyond the walls of the courtroom, reverberating through families, workplaces and friendships, as well as intimate relationships. From the courtroom to the streets of Avignon and Mazan, a team of anthropologists set out to follow the proceedings. Their investigation gathered hundreds of testimonies — from journalists, lawyers, feminist activists, students, healthcare workers, shopkeepers, and many others — offering an unprecedented perspective on this event. This book captures the echoes of a trial that has become a mirror for society, revealing dynamics of power and the relationships between women and men. Written in a clear and compelling style, it invites readers to reflect on one of the most significant moments in the recent history of the struggle against sexual violence. Fourteen social science researchers, all specialists in gender studies, contributed to this book. (Book cover)Septembre 2024. Le procès de Mazan fait l’effet d’une onde de choc. L’événement judiciaire déborde de l’enceinte du tribunal et résonne dans les familles, les quotidiens professionnels, les relations amicales et amoureuses. Depuis la salle d’audience jusque dans les rues d’Avignon et de Mazan, des anthropologues ont voulu suivre ce procès. Leur enquête a permis de recueillir des centaines de témoignages – de journalistes, avocates, membres de collectifs féministes, étudiantes, personnels de santé, commerçantes… – offrant un point de vue inédit sur cet événement. Ce livre donne à entendre les échos d’un procès de société et ce qu’ils révèlent des rapports de pouvoir et des relations entre les femmes et les hommes. Un récit incarné, porté par une plume efficace et fluide, qui invite à penser un des moments les plus importants de l’histoire récente de la lutte contre les violences sexuelles. Quatorze chercheur·es en sciences sociales, spécialistes en études de genre, ont participé à cet ouvrage. Le travail d’enquête et d’écriture a été coordonné par Perrine Lachenal et Céline Lesourd (CNRS/Centre Norbert Elias UMR8562). (4e de couverture
