University of Nice Sophia Antipolis
  • Nice, Provence-Alpes-Cote d'Azur, France
Recent publications
Machine Unlearning (MU) is an emerging discipline studying methods to remove the effect of a data instance on the parameters of a trained model. Federated Unlearning (FU) extends MU to unlearn the contribution of a dataset provided by a client wishing to drop from a federated learning study. Due to the emerging nature of FU, a practical assessment of the effectiveness of the currently available approaches in complex medical imaging tasks has not been studied so far. In this work, we propose the first in-depth study of FU in medical imaging, with a focus on collaborative prostate segmentation from multi-centric MRI dataset. We first verify the unlearning capabilities of a panel of FU methods from the state-of-the-art, including approaches based on model adaptation, differential privacy, and adaptive retraining. For each method, we quantify their unlearning effectiveness and computational cost as compared to the baseline retraining of a model from scratch after client dropout. Our work highlights a new perspective for the practical implementation of data regulations in collaborative medical imaging applications.
Resilience is a key component of system safety evaluation and optimization, and research on natural gas pipeline network system (NGPNS) resilience indices and corresponding evaluation is still in early stages. To evaluate the resilience of NGPNS more synthetically and realistically, and to take into account different forms of disruptions, an integrated simulation model combining the topology and operational parameters is provided. The properties of deterministic disruptions, such as earthquakes and equipment breakdowns, are investigated. The maximum flow method and the shortest path method are combined with operational and structural parameters, to assess the amounts and routes of gas supply before and after disturbance; using complex networks theory and graph theory, the traditional view point is changed from the entire system to the affected area. The results can help guide NGPNS topological design and the development of prewarning schemes, including spare gas sources and gas route optimization, as well as pipeline maintenance strategy. They can also aid in the rapid analysis of disturbance consequences and the improvement of NGPNS resilience evaluating accuracy.
Background Gastric staple line leak treatment after laparoscopic sleeve gastrectomy (LSG) remains challenging. Regenerative medicine is gaining place in the accelerated treatment of damaged tissues. This study presents the first series of gastric leak treatment after LSG using endoscopic intragastric administration of combined autologous mesenchymal stem cells (MSC) and platelet-rich plasma (PRP). Methods MSC-PRP harvesting and endoscopic administration techniques are described in detail. Data were prospectively gathered and analyzed. Primary endpoints were morbidity/mortality rates and fistula closure time. Results Twelve patients (9 women, 3 men) were included. Median age was 41.5 years, median weight 105.5 kg and median BMI 38.9 kg/m². Median time to gastric staple line leak detection was 10 days post-LSG. Median time between re-laparoscopy and MSC-PRP administration was 5 days. MSC-PRP endoscopic administration was successfully performed and tolerated by all patients, with median procedure duration of 27 min and minimal blood loss. Four postoperative complications were noted: two patients with increased tibial pain at tibial puncture site, one with tibial hematoma, and one with epigastric pain/dysphagia. Median length of hospital stay was 1 day. Gastric leak healing occurred after a median of 14 days, only two patients requiring a second MSC-PRP endoscopic injection. Median follow-up was 19 months, all patients being in good health at last contact. Conclusion Endoscopic administration of combined autologous MSC-PRP seems to be a good option for treatment of gastric leaks after sleeve gastrectomy. It is a challenging procedure that should be performed in specialized bariatric centers by expert bariatric surgeons and endoscopists after meticulous patient selection. Graphical Abstract
Recently, numerous physical attacks have been demonstrated against lattice-based schemes, often exploiting their unique properties such as the reliance on Gaussian distributions, rejection sampling and FFT-based polynomial multiplication. As the call for concrete implementations and deployment of postquantum cryptography becomes more pressing, protecting against those attacks is an important problem. However, few countermeasures have been proposed so far. In particular, masking has been applied to the decryption procedure of some lattice-based encryption schemes, but the much more difficult case of signatures (which are highly nonlinear and typically involve randomness) has not been considered until now. In this paper, we describe the first masked implementation of a lattice-based signature scheme. Since masking Gaussian sampling and other procedures involving contrived probability distributions would be prohibitively inefficient, we focus on the GLP scheme of Güneysu, Lyubashevsky and Pöppelmann (CHES 2012). We show how to provably mask it in the Ishai–Sahai–Wagner model (CRYPTO 2003) at any order in a relatively efficient manner, using extensions of the techniques of Coron et al. for converting between arithmetic and Boolean masking. Our proof relies on a mild generalization of probing security that supports the notion of public outputs. We also provide a proof-of-concept implementation to assess the efficiency of the proposed countermeasure.
In 1965, I published a paper, exhibiting a hitherto unknown limit of the Lorentz group, which I christened “Carroll group” due to its seemingly paradoxical physical contents. Since I saw it as more curious than relevant, I published it in French in a journal somewhat afar from the mainstream of theoretical physics at that time. It was most gratifying to witness the quite unexpected favour this paper started to enjoy half a century later, so much that a so-called “Carrollian physics” is now developing, with applications in various domains of forefront theoretical physics, such as quantum gravitation, supersymmetry, string theory, etc. I offer this narrative as an example of the very diverse time scales with which scientific ideas may develop — or not.
This paper studies the dual interaction between economic growth and environmental quality in an endogenous growth model. We exhibit multiple equilibria and complex local and global dynamics, resulting in potential indeterminacy, hysteresis effects, or long-lasting growth and environmental cycles. From a policy perspective, we reveal that changes in the environmental policy should be handled with care, as they may generate aggregate instability or condemn the economy to an environmental poverty trap associated with a possible irreversibility of environmental degradation. Lastly, our analysis provides a reassessment of pollution taxes, which are found to improve long-run economic growth when the model is well-determined, but reduce it in the presence of indeterminacy.
Data-flow reversal is at the heart of source-transformation reverse algorithmic differentiation (reverse ST-AD), arguably the most efficient way to obtain gradients of numerical models. However, when the model implementation language uses garbage collection (GC), for instance in Java or Python, the notion of address that is needed for data-flow reversal disappears. Moreover, GC is asynchronous and does not appear explicitly in the source. This paper presents an extension to the model of reverse ST-AD suitable for a language with GC. The approach is validated on a Java implementation of a simple Navier-Stokes solver. Performance is compared with existing AD tools ADOL-C and Tapenade on an equivalent implementation in C and Fortran.
We tested the hypothesis that breathing heliox, to attenuate the mechanical constraints accompanying the decline in pulmonary function with aging, improves exercise performance. Fourteen endurance-trained older men (67.9 ± 5.9 year, \(\dot{V}\)O2max: 50.8 ± 5.8 ml/kg/min; 151% predicted) completed two cycling 5-km time trials while breathing room air (i.e., 21% O2–79% N2) or heliox (i.e., 21% O2–79% He). Maximal flow–volume curves (MFVC) were determined pre-exercise to characterize expiratory flow limitation (EFL, % tidal volume intersecting the MFVC). Respiratory muscle force development was indirectly determined as the product of the time integral of inspiratory and expiratory mouth pressure (∫Pmouth) and breathing frequency. Maximal inspiratory and expiratory pressure maneuvers were performed pre-exercise and post-exercise to estimate respiratory muscle fatigue. Exercise performance time improved (527.6 ± 38 vs. 531.3 ± 36.9 s; P = 0.017), and respiratory muscle force development decreased during inspiration (− 22.8 ± 11.6%, P < 0.001) and expiration (− 10.8 ± 11.4%, P = 0.003) with heliox compared with room air. EFL tended to be lower with heliox (22 ± 23 vs. 30 ± 23% tidal volume; P = 0.054). Minute ventilation normalized to CO2 production (\(\dot{V}\)E/\(\dot{V}\)CO2) increased with heliox (28.6 ± 2.7 vs. 25.1 ± 1.8; P < 0.001). A reduction in MIP and MEP was observed post-exercise vs. pre-exercise but was not different between conditions. Breathing heliox has a limited effect on performance during a 5-km time trial in master athletes despite a reduction in respiratory muscle force development.
We propose a data monetization architecture based on the Substrate blockchain framework for leveraging automotive radar data. The architecture is designed to enable a virtuous economic cycle involving the Radar automotive data for the consortium members of Automotive enterprises, Vehicle owners, and Radar Equipment manufacturers. The data represented as Non-Fungible Token (NFT) is sourced from the vehicle owners by the Radar component manufacturers and enhanced upon them. It is offered as a token road signature service (NFT) back to the vehicle owners completing the virtuous economic loop. All the participant interactions are monetized and incentivized by dynamic pricing and commission sharing. Data interactions are certified for integrity, and reviews for each NFT are maintained in the blockchain. Architecture respects the fairness, privacy concerns, as well as fidelity aspects. Architecture is implemented in the Substrate Blockchain framework and tested for hybrid consensus scalability of Proof of Authority algorithms. The algorithms evaluated are Aura and BABE, along with GRANDPA for block authoring and finalization, respectively, in a cloud-based implementation that suits the enterprise consortium networks.
We introduce a theoretical and computational framework to use discrete Morse theory as an efficient preprocessing in order to compute zigzag persistent homology. From a zigzag filtration of complexes \((X_i)\), we introduce a zigzag Morse filtration whose complexes \((\mathcal {A}_i)\) are Morse reductions of the original complexes \((X_i)\), and we prove that they both have same persistent homology. This zigzag Morse filtration generalizes the filtered Morse complex of Mischaikow and Nanda Mischaikow and Nanda (Discrete Comput Geom 50(2):330–353, 2013), defined for standard persistence. The maps in the zigzag Morse filtration are forward and backward inclusions, as is standard in zigzag persistence, as well as a new type of map inducing non trivial changes in the boundary operator of the Morse complex. We study in details this last map, and design algorithms to compute the update both at the complex level and at the homology matrix level when computing zigzag persistence. The key point of our construction is that it does not require any knowledge of past and future maps of the input filtration. We deduce an algorithm to compute the zigzag persistence of a filtration that depends mostly on the number of critical cells of the complexes, and show experimentally that it performs better in practice.
A classical planar cable model has been presented in the Irvine textbook and is being used, for example, for the modeling of cable-driven parallel robot. It provides 2 equations involving parameters of the cable material, namely its Young modulus and its linear density, assumed here to be known, and 5 physical parameters namely the coordinates of one cable end-point B, the horizontal and vertical components of the force exerted at B and the length at rest of the cable. This model is extensively used in the modeling of devices where cables are involved (e.g. the kinematics of cable-driven parallel robots) and the model analysis requires to be able to solve very efficiently this 2-equations system when it has 1 or 2 unknowns. We consider various cases where n physical parameters are known and we investigate how the Irvine equations may be exploited to compute the \(5-n\) remaining parameters. In some cases where \(n<3\)it is possible to determine a closed-form solution for this \(5-n\) parameters. But if \(n=3\) it appears that only a numerical approach may allow to get the 2 remaining parameters. We present here a generic algorithm based on a mix of neural networks and deterministic algorithms allows one to get exact solution in that case.
The autocorrelation of a sequence is a useful criterion, among all, of resistance to cryptographic attacks. The behavior of the autocorrelations of random Boolean functions (studied by Rodier et al., (Crypt. Commun. 15, 995–1009, 2023) shows that they are concentrated around a point. We show that the same is true for the evaluation of the periodic autocorrelations of random binary sequences.
Although animal models have helped to elaborate meaningful hypotheses about the pathophysiology of Sudden Unexpected Death in Epilepsy (SUDEP), specific prevention strategies are still lacking, potentially reflecting the limitations of these models and the intrinsic difficulties of investigating SUDEP. The interpretation of pre‐clinical data and their translation to diagnostic and therapeutic developments in patients thus require a high level of confidence in their relevance to model the human situation. Preclinical models of SUDEP models are heterogeneous and range from rodent and non‐rodent species. A critical aspect is whether the animals have isolated seizures exclusively induced by a specific trigger, such models where seizures are elicited by electrical stimulation or pharmacological intervention or DBA mouse strains, or whether they suffer from epilepsy with spontaneous seizures, with or without spontaneous SUDEP, either of non‐genetic epilepsy etiology or of genetically‐based developmental and epileptic encephalopathies. All these models have advantages and potential disadvantages, but it is important to be aware of these limitations to interpret data appropriately in a translational perspective. The majority of models with spontaneous seizures are of a genetic basis whereas SUDEP cases with a genetic basis represent only a small proportion of the total number. In almost all models, cardio‐respiratory arrest occurs during the course of the seizure, contrary to that in patients observed at the time of death, potentially raising the issue of whether we are studying models of SUDEP or models of peri‐seizure death. However, some of these limitations are impossible to avoid and can in part be dependent on specific features of SUDEP, which may be difficult to model. Several preclinical tools are available to address certain gaps in SUDEP pathophysiology, which can be used to further validate current preclinical models.
Background Despite cefoxitin's in vitro resistance to hydrolysis by extended-spectrum beta-lactamases (ESBL), treatment of ESBL-producing Klebsiella pneumoniae (KP) infections with cefoxitin remains controversial. The aim of our study was to compare the clinical efficacy of cefoxitin as definitive antibiotic therapy for patients with ESBL-KP bacteremia in intensive care unit, versus carbapenem therapy. Methods This retrospective study included all patients with monomicrobial bacteremia hospitalized in intensive care unit between January 2013 and January 2023 at the University Hospital of Guadeloupe. The primary outcome was the 30-day clinical success defined as a composite endpoint: 30-day survival, absence of relapse and no change of antibiotic therapy. Cox regression including a propensity score (PS) and PS-based matched analysis were performed for endpoint analysis. Results A total of 110 patients with bloodstream infections were enrolled. Sixty-three patients (57%) received definitive antibiotic therapy with cefoxitin, while forty-seven (43%) were treated with carbapenems. 30-day clinical success was not significantly different between patients treated with cefoxitin (57%) and carbapenems (53%, p = 0.823). PS-adjusted and PS-matched analysis confirmed these findings. Change of definitive antibiotic therapy was more frequent in the cefoxitin group (17% vs. 0%, p = 0.002). No significant differences were observed for the other secondary endpoints. The acquisition of carbapenem-resistant Pseudomonas aeruginosa was significantly higher in patients receiving carbapenem therapy (5% vs. 23%, p = 0.007). Conclusions Our results suggest that cefoxitin as definitive antibiotic therapy could be a therapeutic option for some ESBL-KP bacteremia, sparing carbapenems and reducing the selection of carbapenem-resistant Pseudomonas aeruginosa strains.
Introduction. Despite the availability of effective biologic therapies for psoriasis, there is no gold-standard treatment for nonpustular palmoplantar psoriasis (ppPsO). Methods. G-PLUS, a phase IIIb, double-blind, placebo-controlled, multicentre clinical trial, randomised adults with moderate-to-severe nonpustular ppPsO and limited plaque psoriasis (Psoriasis Area and Severity Index (PASI) ≥3 but <10) to guselkumab (an interleukin-23p19 blocker) or placebo. Placebo participants were crossed over to receive guselkumab at week (Wk) 16. The primary efficacy endpoint was the proportion of participants achieving palmoplantar PASI (ppPASI) 75 response at Wk16; clinical, biomarker, and quality-of-life endpoints were assessed through Wk48 and safety through Wk56. Results. At Wk16, ppPASI75 response was achieved by 35.9% of the guselkumab participants compared with 28.2% in the placebo group, resulting in a 7.7% difference in response rates (95% confidence interval: −11.5 and 24.7), which was not statistically significant ( p = 0.533 ). More pronounced numerical improvements favouring guselkumab were observed for more stringent efficacy endpoints, such as Wk16 palmoplantar Investigator’s Global Assessment (ppIGA) 0/1 response (guselkumab 34.6% vs. placebo 15.4%). Through Wk48, further improvements were observed in ppPASI75 response (55.1% and 64.1%) and ppIGA 0/1 response (42.3% and 48.7%) for the guselkumab and placebo-crossover groups, respectively. Dermatology Life Quality Index responses showed comparable trends at both timepoints. Safety and pharmacodynamic findings were consistent with the established profile for guselkumab. Serum biomarker levels were significantly reduced with guselkumab and correlated with the baseline PASI score but not the ppPASI score. Conclusion. Although the primary endpoint was not met, analysis of stringent secondary endpoints and post hoc analyses showed numerical improvements favouring guselkumab at Wk16. There were no new safety signals. Further studies are warranted to better understand the impact of guselkumab treatment in patients with ppPsO. This trial is registered with NCT03998683.
Establishing equivalence and refinement relations between programs is an important mean for verifying their correctness. By establishing that the behaviours of a modified program simulate those of the source one, simulation relations formalise the desired relationship between a specification and an implementation, two equivalent implementations, or a program and its optimised implementation. This article discusses a notion of simulation between open automata, which are symbolic behavioural models for communicating systems. Open automata may have holes modelling elements of their context, and can be composed by instantiation of the holes. This allows for a compositional approach for verification of their behaviour. We define a simulation between open automata that may or may not have the same holes, and show under which conditions these refinements are preserved by composition of open automata.
Human knowledge is subject to uncertainties, imprecision, incompleteness and inconsistencies. Moreover, the meaning of many everyday terms is dependent on the context. That poses a huge challenge for the Semantic Web. This paper introduces work on an intuitive notation and model for defeasible reasoning with imperfect knowledge, and relates it to previous work on argumentation theory. PKN is to N3 as defeasible reasoning is to deductive logic. Further work is needed on an intuitive syntax for describing reasoning strategies and tactics in declarative terms, drawing upon the AIF ontology for inspiration. The paper closes with observations on symbolic approaches in the era of large language models.
Background: ALK, ROS1 and RET rearrangements occur respectively in 5%, 2% and 1% non-small cell lung cancers (NSCLC). ALK and ROS1 fusion proteins detection by immunohistochemistry (IHC) has been validated for rapid patient screening, but ROS1 fusions need to be confirmed by another technique and no RET IHC test is available for clinical use. Research design and methods: We report herein the usefulness of the HTG EdgeSeq Assay, an RNA extraction-free test combining a quantitative nuclease protection assay with NGS, for the detection of ALK, ROS1 and RET fusions from 'real-life' small NSCLC samples. A total of 203 FFPE samples were collected from 11 centers. They included 143 rearranged NSCLC (87 ALK, 39 ROS1, 17 RET) and 60 ALK-ROS1-RET negative controls. Results: The assay had a specificity of 98% and a sensitivity for ALK, ROS1 and RET fusions of 80%, 94% and 100% respectively. Among the 19 HTG-assay false negative samples, the preanalytical conditions were identified as the major factors impacting the assay efficiency. Conclusions: Overall, the HTG EdgeSeq assay offers comparable sensitivities and specificity than other RNA sequencing techniques, with the advantage that it can be used on very small and old samples collected multicentrically.
Institution pages aggregate content on ResearchGate related to an institution. The members listed on this page have self-identified as being affiliated with this institution. Publications listed on this page were identified by our algorithms as relating to this institution. This page was not created or approved by the institution. If you represent an institution and have questions about these pages or wish to report inaccurate content, you can contact us here.
3,928 members
Philippe R Franken
  • Faculty of Medicine
Bernard Moussian
  • Institut Sophia Agrobiotech (UMR ISA 1355 INRA / UNS / 7254 CNRS)
Pascal Staccini
  • Risk Engineering and Medical Informatics
28 avenue de Valrose, 06103, Nice, Provence-Alpes-Cote d'Azur, France
Head of institution
Pr. Frédérique Vidal
+33 (0)4 92 07 60 60