HAL-Université de Bretagne Occidentale
Not a member yet
75442 research outputs found
Sort by
Assistance opératoire par simulateurs chirurgicaux et intelligence artificielle
No full textCataract surgery is the only curative treatment for this disease and the most common surgical procedure in ophthalmology. With an aging population, the demand for skilled surgeons is increasing, requiring solutions to enhance the training and integration of future ophthalmologists. A digital mentor capable of detecting, or even anticipating, surgical errors in real-time could be a valuable aid in supporting less experienced surgeons. However, predicting surgical errors remains a challenge, mainly due to the lack of annotated data with surgical errors. This thesis explores the use of a surgical simulator, employed in the initial training of surgeons, to train deep learning models able to predict the simulator’s assessments and identifyingsurgical errors from kinematic and video data, leveraging the errors already present in the simulator data. Two modules were studied: navigation (tremor reduction) and capsulorhexis (a critical step in surgery). The results are promising and pave the way for surgical error anticipation. To validate these approaches in a real clinical setting, knowledge transfer between simulated and real surgery was explored through unsupervised domain adaptation. A third dataset of capsulorhexis from real surgeries was specifically collected for this study. To our knowledge, this work is the first initiative for anticipating surgical errors in cataract surgery.La chirurgie de la cataracte est le seul traitement curatif de cette maladie et constitue l’intervention chirurgicale la plus courante en ophtalmologie. Face au vieillissement de la population, la demande en chirurgiens qualifiés augmente, nécessitant des solutions pour améliorer la formation et l’intégration des ophtalmologistes. Un mentor numérique capable de détecter, voire d’anticiper les erreurs chirurgicales en temps réel pourrait constituer une aide précieuse pour accompagner les chirurgiens peu expérimentés. Cependant, la prédiction des erreurs chirurgicales reste un défi, principalement en raison du manque de données annotées incluant des erreurs. Cette thèse explore l’utilisation d’un simulateur chirurgical, utilisé pour la formation initiale des chirurgiens, pour entrainer des modèles d’apprentissage profond capables de prédire les évaluations du simulateur et d’identifier les erreurs chirurgicales à partir de données cinématiques et vidéo, en tirant parti des erreurs présentes dans les données du simulateur. Deux modules ont été étudiés : la navigation (réduction des tremblements) et le capsulorhexis (étape cruciale de la chirurgie). Les résultats obtenus sont encourageants et ouvrent la voie à l’anticipation des erreurs chirurgicales. Afin de valider ces approches dans un contexte clinique réel, un transfert de connaissances entre chirurgie simulée et réelle a été exploré via une adaptation de domaine non supervisée. Une troisième base de données de capsulorhexis, issus de la chirurgie réelle, a ainsi été rassemblée spécialement pour cette étude. Ce travail constitue la première initiative d’anticipation des erreurs chirurgicales dans la chirurgie de la cataracte
Corsaires, cargaisons et droit maritime pendant la Révolution française
No full textInternational audienc
Affreux, sales… et malades ? La campagne navale française du printemps 1794 au prisme de l’hygiène et de la santé
No full textInternational audienc
Wide-band multifrequency MR elastography with a fractional viscoelastic model and nonlinear inversion for enhanced viscoelastic parameter mapping
No full textInternational audienc
Typologie des habitats benthiques de Saint-Pierre-et-Miquelon
No full textCe rapport présente la liste hiérarchisée des habitats benthiques de Saint-Pierre et Miquelon et leurs fiches descriptives illustrées. Ce travail est le fruit d’une collaboration initiée en 2023 entre le CNRS (LEMAR) et PatriNat (OFB - MNHN). Les habitats inventoriés ont majoritairement été identifiés et décrits grâce à deux missions scientifiques réalisées entre 2023 et 2025. Plus de 170 sites, répartis tout autour de l’archipel, à différentes profondeurs, sur différents types de fonds et avec différents niveaux d’exposition à la houle ont été prospectés afin de maximiser la diversité d’habitats marins rencontrés. Au total, la typologie regroupe 141 unités avec quatre niveaux de hiérarchisation :1 - l’étage (du supralittoral au bathyal)2 - le substrat (de la roche aux sables fins)3 - l’habitat (49 unités, chacune accompagnée d’une fiche descriptive)4 - les variantes de l’habitat (59 unités).Ce document comprend également une présentation du territoire et des grands types d’habitats benthiques présents à Saint-Pierre-et-Miquelon, un schéma de l’étagement ainsi que trois transects illustrant la succession typique d’habitats de la côte vers le large dans différents secteurs
Fusion de données AVQ pour la contextualisation du coach virtuel e-VITA
No full textInternational audienceFusion de données AVQ pour la contextualisation du coach virtuel e-VIT
Surface modification of extracellular vesicles with polyoxazolines to enhance their plasma stability and tumor accumulation
No full textInternational audienceExtracellular vesicles (EVs) are future promising therapeutics, but their instability in vivo after administration remains an important barrier to their further development. Many groups evaluated EV surface modification strategies to add a targeting group with the aim of controlling EV biodistribution. Conversely, fewer groups focused on their stabilization to obtain "stealth" allogenic EVs. Modulating their stabilization and biodistribution is an essential prerequisite for their development as nano-therapeutics. Here, we explored polyoxazolines with lipid anchors association to the EV membrane (POxylation as an alternative to PEGylation) to stabilize EVs in plasma and control their biodistribution, while preserving their native properties. We found that this modification maintained and seemed to potentiate the immunomodulatory properties of EVs derived from mesenchymal stem/stromal cells (MSC). Using a radiolabeling protocol to track EVs at a therapeutically relevant concentration in vivo, we demonstrated that POxylation is a promising option to stabilize EVs in plasma because it increased EV half-life by 6 fold at 6 h post-injection. Moreover, EV accumulation in tumors was higher after POxylation than after PEGylation
Detection of Seismic and Acoustic Sources Using Distributed Acoustic Sensing Technology in the Gulf of Catania
No full textInternational audienceDistributed Acoustic Sensing (DAS) technology presents an innovative method for marine monitoring by adapting existing underwater optical fiber networks. This paper examines the use of DAS with the Istituto Nazionale di Fisica Nucleare–Laboratori Nazionali del Sud (INFN-LNS) optical fiber infrastructure in the Gulf of Catania, Eastern Sicily, Italy. This region in the Western Ionian Sea provides a unique natural laboratory due to its tectonic and volcanic activity, proximity to Mount Etna, diverse marine ecosystems and significant human influence through maritime traffic. By connecting a 28 km long optical cable to an Alcatel Submarine Network OptoDAS interrogator, DAS successfully detected a range of natural and human–made signals, including a magnitude 3.5 ML earthquake recorded on 14 November 2023, and acoustic signatures from vessel noise. The earthquake–induced Power Spectral Density (PSD) increased to up to 30 dB above background levels in the 1–15 Hz frequency range, while vessel noise exhibited PSD peaks between 30 and 60 Hz with increases of up to 5 dB. These observations offered a detailed spatial and temporal resolution for monitoring seismic wave propagation and vessel acoustic noise. The results underscore DAS’s capability as a robust tool for the continuous monitoring of the rich underwater environments in the Gulf of Catania
Late Holocene linkages between soil erosion, phosphate bioavailability and lake productivity in Central Africa
No full textInternational audienceMajor vegetation shifts and large-scale erosion events occurred in Central Africa during the late Holocene in response to hydroclimate change and enhanced land use related to the Bantu expansion. On the Adamawa Plateau, a highland in Central Cameroon that experienced abrupt forest collapse about 3000 years ago, recent works showed that lake primary productivity markedly increased during the same period. Yet, the causes of increased productivity in Adamawa plateau lakes and its link to the Central African rainforest crisis still remain unclear. To address this issue, we investigated lake sediment records from this region, using radiogenic isotope proxies for sediment provenance and chemical weathering to complement previously published mineralogical and organic geochemistry data. Our results indicate that increased biological productivity in Adamawa Plateau lakes was linked to enhanced erosion of highly weathered lateritic soils. An examination into possible causes leads us to propose that phosphate desorption following erosion of kaolinite, iron (goethite) and aluminium (gibbsite) oxide-bearing soils may have played a key role in stimulating lake productivity at that time, together with degradation of terrestrial organic matter. These findings emphasize the importance of deforestation and land use on soil-P cycling and lake ecosystems in tropical regions during the Holocene
