Institute for Radiation Protection and Nuclear Safety (IRSN)
Not a member yet
7928 research outputs found
Sort by
Artificial Intelligence and generative data augmentation for automated chromosomal aberration detection in cytogenetic FISH imaging
No full textInternational audienceFollowing accidental exposure to ionizing radiation, it is necessary to refine the assessment of the dose received to conduct effective sorting of asymptomatic victims. Among the available dosimetric techniques, biological dosimetry based on cytogenetic imaging consists of counting chromosomal aberrations (CA) within circulating lymphocytes. These aberrations can be unstable (ex. dicentrics) or more persistent (ex. translocations). The latter are more suitable to perform retrospective dosimetry several years after exposure and can be identified through color colocalizations in fluorescence in situ hybridization (FISH) imaging.The counting procedure is long and tedious, requiring trained biologists and to our knowledge, no automated solutions are presently available. The present study aims to develop a deep learning-based workflow for detection of stable CA in FISH images. Thanks to recent advances in computer vision, we deployed state-of-the-art object detection models allowing the localization and classification of fluorescent chromosomes. We faced two major challenges: the limited number of annotated data and the rarity of the translocation at the metaphase scale. The idea was then to develop a semi-supervised approach taking advantage of both spatial annotations and unannotated data.Faster-RCNN, an object detection neural network, is relevant here because it can simultaneously locate and classify chromosomes. Additionally, we studied classification at the chromosome level via ResNet models, after segmentation by U-Net models. Both approaches lead to accurate detection of fluorescent chromosomes, but the distinction of aberrations remains subject to improvement.To overcome the restriction in the number of annotated data, we explored the possibility of creating synthetic data using generative diffusion models. More specifically, image-to-image models can transform an input image to match the characteristics of a target image. In this work, we generated artificially translocated chromosomes from the blue channels (DAPI equivalent) of our unannotated images.In conclusion, the present work provides promising results about deep learning model deployment for an automated chromosomal aberration detection in cytogenetic imaging
Main Outcomes for AAR Studies on Large Scale Experiments in the ODOBA International Project
No full textInternational audienceConsidering the importance of the potential effects induced by internal swelling reaction (ISR) on the third barrier of confinement for nuclear power plants and the lack of knowledge on ISR development on massive structure, IRSN launched the ODOBA international project in 2016. Indeed, the knowledge on ISR relies mainly on small-scale and separate effect experiments in laboratories, or medium scale outdoor experiments. Their kinetics and the structure effect of the different parameters involved (temperature, hygrometry, mix, reinforcement…) depend on the studied scale. The ODOBA project is based on experiments on very large concrete blocks widely instrumented, located in Cadarache (France). The concrete mixes are chosen with the objective of developing ISR: either the delayed ettringite formation (DEF), or the alkali-aggregate reaction (AAR), or a possible coupling between AAR and DEF. These mixes have the particularity of being as close as possible to those used for French nuclear reactor containment buildings, but with lower alkali level than former similar experiments. Concerning the 20 tons AAR blocks poured from 2016 to 2018, the residual expansion tests on small samples extracted by coring showed a very weak development of AAR pathology, with a swelling qualified as negligible according to the recommendations in force. An overview of the state of the blocks devoted to AAR is introduced by successively presenting the formulations of the concretes, their characterizations by destructive tests and their follow-up from early age until now without ageing acceleration. The lessons learned give a better assessment of AAR development possibilities for a mix using medium alkali level and classical Potentially Reactive aggregates at very large scale
Highlighting cross section library effects on absorption rates and flux for natural elements in a thermal spectrum using TRIPOLI-5
No full textInternational audienceThis work presents an investigation of cross section library effects with the new TRIPOLI-5 Monte Carlo code, a joint project between CEA, IRSN and EDF in France. Cross section library effects on the absorption rate and the flux spectrum were highlighted for natural elements subjected to a thermal neutron flux. The numerical benchmark that is presented here is a sphere with a radius of 30 cm filled with 10 -2 atoms•(barn•cm) -1 of a single element according to its natural isotopic abundances. For the thermal flux case, a smaller sphere with a radius of 1 cm is placed at the center of the main sphere to obtain a thermal-like flux from a pointwise Maxwellian source using a mixture of hydrogen and boron-10. A total of 12 cross sections libraries have been compared
Compartment fires: Challenges for fire modeling as a tool for a safe design (IAFSS workshop, April 2021)
No full textInternational audienceThe use of fire models to support fire protection engineering decisions requires an understanding of model shortcomings and assurance in their predictive robustness. This note is a summary of the online ‘compartment fire’ workshop that was organized prior to the International Association of Fire Safety Science (IAFSS) symposium, hosted online by the University of Waterloo (Canada) in April 2021. The objectives of the Workshop were to identify, discuss and prioritize key compartment fire modeling challenges.It is recognized that the substantial changes in the built environment and the variety of subsequent fire dynamics problems necessitate significant advances in modelling, particularly with respect to (amongst other aspects) (i) under-ventilated fires (the prediction of soot and CO concentrations, extinction and reignition), (ii) heat transfer and the interaction with structural and non-structural elements, and (iii) the interaction with water-based fire suppression systems. High-quality and well-documented experimental tests play an essential role in fostering model development and validation; a good synergy between experimentalists and modellers is of utmost importance
Conditional probability of surface rupture: A numerical approach for principal faulting
No full textInternational audienceThis study presents a numerical approach for probabilistic fault displacement hazard analysis (PFDHA), aimed at addressing an alternative solution with commonly used empirical methodologies. Our model utilizes probability distributions to compute the conditional probability of surface rupture (CPSR). Leveraging earthquake catalogs, we derived the hypocentral depth distribution (HDD) across eight globally distributed seismotectonic regions categorized by faulting kinematics (normal, reverse, strike-slip). We calculated the hypocentral depth ratio (HDR) distribution, to model rupture position from the hypocenter. Employing magnitude scaling relations we determined rupture widths (W) spanning magnitudes 5–8. User-input parameters, including fault style, average dip angle, and seismogenic depth, with associated uncertainties, derive the CPSR estimation of surface rupture occurrences. Our findings highlight seismogenic depth as the most influential parameter and reveal correspondences between empirical curves derived for specific regions, emphasizing the importance of site-specific rupture probability assessments over global datasets and underscores the significance of considering seismotectonic context when evaluating fault displacement hazard. The numerical code for CPSR calculation has been developed and is openly accessible on GitHub
First Steps of the ANTARES Validation with TVA Watts Bar Unit 1 Cycle 1 Benchmark
No full textInternational audienceIn recent years, IRSN has developed a new project to simulate accurately the interaction between neutronic and thermal-hydraulic phenomena. This project called ANTARES (Advanced Neutronics and Thermal-hydraulic for the Analysis of the Reactor Safety) couples the thermal-hydraulic code CATHARE-3 with the 3D neutronic nodal code PARCS. The aim of this project is to provide IRSN with a multi-physics computing tool for carrying out safety analyses primarily for PWRs. To increase the reliability of ANTARES, IRSN has decided to participate in the "TVA Watts Bar Unit 1 Multi-Physics Multi-Cycle Depletion" benchmark proposed and supported within the framework of the OECD/NEA/WPRS working group. After a brief description of the ANTARES computational chain, the paper presents our first results, which demonstrate a good agreement between ANTARES and the proposed reference solutions by the benchmark team
Prophylaxis by Administration of Stable Iodine: Principle, Evolution and Recommendations
No full textInternational audienceProphylaxis by stable iodine “KI 65 mg, breakable tablet” is a pharmacological countermeasure of radiological protection adapted for the prevention of thyroid cancer after exposure to radioactive iodine. A first modification of the marketing authorization of “KI 65 mg, breakable tablet” for repeated prophylaxis for adults and children above 12 years old was obtained in France in 2020. At this point in time, the recommendation of a unique “KI” intake by pregnant and breastfeeding women and children below 12 years old remain valid
Aerosol characterization on noisy surfaces by deep-learning methods
No full textInternational audienceExperimental studies to characterize aerosol depositionin flow system (such as ventilation duct) may requireavoiding a sampling substrate which modifies the flow inthe boundary layer, influencing the deposition or theinteraction forces. However, measuring the deposition ofaerosols on surface without any intermediate substratepresents detection difficulties (Costa et al. 2021) whichcan be overcome by direct visualization of aerosols bymicroscopy, using the microscope as a scanner.However, analysing the deposition of aerosols on anysurfaces by microscopy generates certain constraintsregarding the use of known visualization techniques:inhomogeneous image backgrounds due to the surfacestructure, unsharp images due to the non-flatness of thereal considered surfaces. Common image processing isthen not very effective.The objective of this work is therefore to deployadvanced computer vision methods to process imagesacquired on typical surfaces of ventilation ducts(stainless steel).Little work exists to date on the detection of aerosolsusing neural networks. Monchot et al. (2021) wereinterested in nanoparticles, as were Merouane et al.(2023) but in these works, the particles stand outrelatively well from the background
Analyse dose-réponse à partir d'imagerie cérébrale après radiothérapie via un mélange d'expert spatial
No full textNational audienceRadiotherapy (RT) is one of the most important treatments for brain tumors. However, its potential toxicity on the central nervous system is a highly relevant clinical issue as cognitive dysfunction, mainly related to radiation-induced leukoencephalopathy, may alter the quality of life of patients. As part of the RADIO-AIDE research project, the aim of this work is to model and learn about the potential relationship between the dose of ionizing radiation absorbed in a voxel of the brain after RT (from CT-scan images) and the presence/absence of brain lesions in these voxels (from segmented brain MRI data). We propose to extend the class of mixture of experts’ models, including the well-known Gaussian Locally Linear Mapping model (GLliM), to a binary outcome and a spatially structured predictor. We thus propose and compare several mixtures of experts’ models based on a piecewise logistic regression and different spatial components (hidden Potts model, conditionally auto-regressive model) to account for dependency between neighboring voxels. Various Bayesian statistical learning methods (variational Bayes, MCMC, SMC) are implemented and compared from simulated data as well as real data from the EpiBrainRad prospective cohort, which includes patients treated with radiochemotherapy for glioblastoma. Many modelling perspectives and Bayesian computational challenges will also be discussed.La radiothérapie (RT) est l'un des traitements les plus importants pour les tumeurs cérébrales. Cependant, sa toxicité potentielle sur le système nerveux central est un des premiers problème clinique, car le dysfonctionnement cognitif, principalement lié à la leuco-encéphalopathie radio-induite, peut altérer la qualité de vie des patients. Dans le cadre du projet de recherche RADIO-AIDE, l’objectif de ce travail est de modéliser et d'estimer la potentielle association entre la dose de rayonnement ionisant absorbée dans un voxel du cerveau après RT (à partir d’images CT-scan) et la présence/absence de lésions cérébrales dans ces voxels (à partir de données IRM cérébrales segmentées). Nous proposons d’étendre la classe de mélange de modèles d’experts, à un résultat binaire et à un prédicteur spatialement structuré. Nous proposons et comparons ainsi plusieurs mélanges de modèles d’experts basés sur une régression logistique par morceaux et différentes composantes spatiales (modèle de Potts caché, modèle auto-régressif conditionnel) pour tenir compte de la dépendance entre voxels voisins. Différentes méthodes d'apprentissage statistique bayésiennes (Bayes variationnel, MCMC, SMC) sont mises en œuvre et comparées à partir de données simulées ainsi que de données réelles issues de la cohorte prospective EpiBrainRad, qui inclut des patients traités par radiochimiothérapie pour glioblastome. De nombreuses perspectives de modélisation et défis informatiques bayésiens seront également abordés
Réseau Caspa - Guideline pour les projets de métrologie participative
No full textCe guideline est dédié à la mise en œuvre d’une démarche de recherche basée sur les métrologies participatives (dispositifs de citizen science mobilisant des capteurs environnementaux). Il a été conçu par une équipe de recherche du réseau CASPA (Capteurs et Sciences Participatives) dans le cadre d’un programme interdisciplinaire financé par la mission MITI (Mission pour les Initiatives Transverses et Interdisciplinaires) du CNRS (2021-2023). Le guide se présente sous la forme d’une aide à la réflexion pour la conception d’une démarche de métrologie participative. Il peut également servir de base pour l'auto-évaluation amont d’un projet ou l'analyse de celui-ci. Le GuideLine Caspa a été réalisé par : Arruabarrena B. et Carmes M., Dicen-IDF/CNAM ; Bottolier JF et Martin R., IRSN, Open Radiation;Kouadio J.S, Institut de Recherche en Sciences et Techniques de la Ville (IRSTV), Université Gustave Eiffel;Schlupp A., Université de Strasbourg, CNRS, ITES UMR 7063;Scotto D'Apollonia L., laboratoire citoyen Artivistes Atelier/LIRDEF Université Montpellierhttps://caspa.fr