HAL Portal Univ-Côtedazur
Not a member yet
95391 research outputs found
Sort by
Real-life data on hepatocellular carcinoma in the French prospective CHIEF cohort in the era of immunotherapy
No full textInternational audienceBackground The burden of hepatocellular carcinoma (HCC) increases worldwide. We report the current landscape of HCC, in France. Methods Carcinome HépatocellulaIrE en France (CHIEF) is a national, prospective, observational cohort initiated in 2019 with the aim of including 5000 patients with HCC, with a 5-year follow-up for each. CHIEF Epidemio 2000, is the first global analysis. Results In September 2021, 2043 patients were included in 32 centers. We analyzed 1640 patients, 86% men, 68-year-old, BMI 26.8. 70.8% of patients had cirrhosis (MELD score 9, Child–Pugh A in 77.8%, and portal hypertension in 39%. Liver disease was related to alcohol 58.5%, metabolic syndrome 39%, and virus 23.3%. HCC was confirmed by histology in 46.3%. The Milan criteria fulfilled 32.9%, median AFP 39 ng/ml, 5.9% portal thrombosis, and 10.7% metastases. BCLC classes 0, A, B, C, and D were 6.1, 29.8, 28.8, 32.1, and 3.2%, respectively. HCC was detected during a surveillance program in 35.2% associated with better survival ( P < 0.001). Median follow-up was 17.76 months (29.1% deaths). The 6, 12, and 18 months overall survival rates were 84.9% (95% CI: 82.8–87), 76.7% (95% CI: 74.2–79.2), 69.3% (95% CI: 66.4–72.3). One-year survival for BCLC 0, A, B, C, and D was at 95.6, 89.7, 81.7, 54.9, and 40%, respectively ( P < 0.0001). First-line treatment was curative, locoregional, or systemic in 40.5, 36.2, and 19.2%, with 1-year survival at 92.9, 82.2, and 57.8%, respectively ( P < 0.0001). Atezolizumab–bevacizumab yielded a median overall survival of 17.05 months versus nine for TKI ( P < 0.0001). Conclusion In real-life data, metabolic syndrome becomes the second cause of HCC in France. The 1-year survival rates are high for all treatments applied, and Immunotherapy yields similar results than in trials
Revisiting the Fast Fourier Transform in Rocq
No full textThis notes explains how a standard algorithm that constructs the discrete Fourier transform has been formalised and proved correct in the Coq proof assistant using the SSReflect extension
Euclid preparation. Angular power spectra from discrete observations
No full textInternational audienceWe present the framework for measuring angular power spectra in the Euclid mission. The observables in galaxy surveys, such as galaxy clustering and cosmic shear, are not continuous fields, but discrete sets of data, obtained only at the positions of galaxies. We show how to compute the angular power spectra of such discrete data sets, without treating observations as maps of an underlying continuous field that is overlaid with a noise component. This formalism allows us to compute exact theoretical expectations for our measured spectra, under a number of assumptions that we track explicitly. In particular, we obtain exact expressions for the additive biases ("shot noise") in angular galaxy clustering and cosmic shear. For efficient practical computations, we introduce a spin-weighted spherical convolution with a well-defined convolution theorem, which allows us to apply exact theoretical predictions to finite-resolution maps, including HEALPix. When validating our methodology, we find that our measurements are biased by less than 1% of their statistical uncertainty in simulations of Euclid's first data release
Euclid preparation. The impact of relativistic redshift-space distortions on two-point clustering statistics from the Euclid wide spectroscopic survey
No full textInternational audienceMeasurements of galaxy clustering are affected by RSD. Peculiar velocities, gravitational lensing, and other light-cone projection effects modify the observed redshifts, fluxes, and sky positions of distant light sources. We determine which of these effects leave a detectable imprint on several 2-point clustering statistics extracted from the EWSS on large scales. We generate 140 mock galaxy catalogues with the survey geometry and selection function of the EWSS and make use of the LIGER method to account for a variable number of relativistic RSD to linear order in the cosmological perturbations. We estimate different 2-point clustering statistics from the mocks and use the likelihood-ratio test to calculate the statistical significance with which the EWSS could reject the null hypothesis that certain relativistic projection effects can be neglected in the theoretical models. We find that the combined effects of lensing magnification and convergence imprint characteristic signatures on several clustering observables. Their S/N ranges between 2.5 and 6 (depending on the adopted summary statistic) for the highest-redshift galaxies in the EWSS. The corresponding feature due to the peculiar velocity of the Sun is measured with a S/N of order one or two. The from the catalogues that include all relativistic effects reject the null hypothesis that RSD are only generated by the variation of the peculiar velocity along the line of sight with a significance of 2.9 standard deviations. As a byproduct of our study, we demonstrate that the mixing-matrix formalism to model finite-volume effects in the can be robustly applied to surveys made of several disconnected patches. Our results indicate that relativistic RSD, the contribution from weak gravitational lensing in particular, cannot be disregarded when modelling 2-point clustering statistics extracted from the EWSS
Print Exposure and Reading Development in the French Educational Context: A Systematic Review
No full textThis systematic review examines the impact of print exposure on reading development within the French education system and its implications for international research. Motivated by France's declining performance in international reading assessments, the review synthesizes evidence on how different forms of print exposure-ranging from home literacy environments to educational interventions-contribute to literacy outcomes. It critically evaluates print exposure measures, ranging from self-reports to print exposure checklists, and considers their predictive validity and conceptual clarity. Consistent with international findings, the review confirms that print exposure, and more broadly the home literacy environment, supports literacy development. However, its role appears more limited in France compared to other OECD countries. This may be attributed to the strong effects of socioeconomic factors on students' performance, particularly pronounced in France, and the centralized educational system, where literacy instruction is primarily school-driven, potentially reducing parental involvement. Building on these insights, we propose an integrative framework to refine the conceptualization of print exposure by distinguishing it from related constructs. Specifically, we suggest separating direct measures of print exposure from distal indicators like the number of books at home, which often overlap with socioeconomic variables. We also advocate for the inclusion of other potentially relevant variables, such as play experiences and oral parent-child interactions. To improve parsimony and robustness, we propose that future research adopt a nested incremental approach to identify a minimal set of variables that maximize both predictive power and reliability
Cross-Modal Neuromorphic Semantic Segmentation based on Knowledge Distillation
No full textInternational audienceEvent-based cameras, with their high temporal resolution and low energy consumption, offer significant advantages over conventional cameras for computer vision tasks. However, current semantic segmentation algorithms for event-based data face challenges in achieving optimal performance for two main reasons: 1) the mismatch between sparse event streams (even when encoded into pseudo-frames) and the dense data structure of traditional frame-based images for which most existing implementations were designed, and 2) the lack of texture information in events, which solely detect temporal variations in brightness.To address this challenge, we propose a novel Cross-Modal (CM) Knowledge Distillation (KD) approach. Our method transfers knowledge from a high-performing Artificial Neural Network (ANN) processing fused grey-scale images and events to a Spiking Neural Network (SNN) -a bio-inspired, energy-efficient computing paradigm -operating on event data alone. Experiments on the DDD17 and DSEC-semantic datasets demonstrate that our approach significantly improves semantic segmentation results while reducing SNN energy consumption. This work represents the first application of cross-modal knowledge distillation to neuromorphic semantic segmentation, paving the way for more efficient event-based vision systems.</p
Euclid preparation. Simulating thousands of Euclid spectroscopic skies
No full textInternational audienceWe present two extensive sets of 3500+1000 simulations of dark matter haloes on the past light cone, and two corresponding sets of simulated (`mock') galaxy catalogues that represent the Euclid spectroscopic sample. The simulations were produced with the latest version of the PINOCCHIO code, and provide the largest, public set of simulated skies. Mock galaxy catalogues were obtained by populating haloes with galaxies using an halo occupation distribution (HOD) model extracted from the Flagship galaxy catalogue provided by Euclid Collaboration. The Geppetto set of 3500 simulated skies was obtained by tiling a 1.2 Gpc/h box to cover a light-cone whose sky footprint is a circle of 30 deg radius, for an area of 2763 deg and a minimum halo mass of Msun/h. The relatively small box size makes this set unfit for measuring very large scales. The EuclidLargeBox set consists of 1000 simulations of 3.38 Gpc/h, with the same mass resolution and a footprint that covers half of the sky, excluding the Milky Way zone of avoidance. From this we produced a set of 1000 EuclidLargeMocks on the 30 deg radius footprint, whose comoving volume is fully contained in the simulation box. We validated the two sets of catalogues by analysing number densities, power spectra, and 2-point correlation functions, showing that the Flagship spectroscopic catalogue is consistent with being one of the realisations of the simulated sets, although we noticed small deviations limited to the quadrupole at k>0.2 h/Mpc. We show cosmological parameter inference from these catalogues and demonstrate that using one realisation of EuclidLargeMocks in place of the Flagship mock produces the same posteriors, to within the expected shift given by sample variance. These simulated skies will be used for the galaxy clustering analysis of Euclid's Data Release 1 (DR1)
CEERS: Possibly forging the first dust grains in the universe: A population of galaxies with spectroscopically derived extremely low dust attenuation (GELDA) at 4.0 < z ≲ 11.4
No full textInternational audienceAims. This work aims to investigate the coevolution of metals and dust for 173 galaxies at 4.0 8.8 (83.3% of the total sample) than at z < 8.8 (26.3% of the total sample). This suggests that GELDAs became dominant in the early Universe. Assuming a prior far-infrared dust spectrum from the ALPINE sample, we performed an analysis of the properties of this galaxy population. The trends observed in the Mdust versus Mstar diagram feature an upper and a lower sequence linked by objects that can be transitional. A comparison with models suggests that we might observe a critical transition at Mstar≈108.5 M⊙, corresponding to a critical metallicity of Zcrit = 12+log10(O/H) ≈ 7.60 (i.e. Z/Z⊙≈0.1). At this point, galaxies transition from being dominated by stellar-dust production (mainly from supernovae) to grain growth through gas–dust accretion in the ISM. The observational critical metallicity Zcrit derived in this paper is in good agreement with predictions from theoretical models for the onset of efficient grain growth. Furthermore, the mean gas-mass fraction of our entire sample at 4.0 < z < 11.4 is very high: fgas≳0.9. All of our galaxies, including GELDAs at all redshifts, contain a large amount of gas that was not expelled from the galaxies. Finally, the small size of the galaxies combined with the mass of gas lead to very high surface-gas densities – which put our sample below high-redshift sub-millimeter galaxies – at relatively low star formation efficiency. The population of high-redshift GELDAs would provide us with a natural and inherent explanation for the origin of the apparent tension between observations and theoretical models in the number density of bright galaxies at z ≳ 9
System Design Using Sensor Compression and Grey Wolf Optimizer for AI-based Security Applications
No full textInternational audienceThe security of public spaces is a major concern. In particular, there is a need for compact and non invasive concealed threat detection systems. In this paper, we investigate the reduction of the number of sensors using a joint design scheme involving the number of antennas and the Support Vector Machine (SVM) classification algorithm. We have developed a joint optimization that combines a bio-inspired optimized method, called Grey Wolf Optimizer, and the Run-Length Encoding (RLE) compression method. Results are compared to a reference case using 22,801 sensors. The classification performance reaches 75%, with specificity and sensitivity (that account for lethal objects) values of 66% and 83%, respectively. The optimized system reduces the number of sensors to 2,818 with an overall detection performance of 79%. A sub-sampling method further reduces the system to 1,038 sensors, achieving 100% sensitivity, guaranteeing detection of all concealed threats