Portail HAL UNIV-RENNES
Not a member yet
180610 research outputs found
Sort by
How to organise a scientific competition to benchmark methods and algorithms in computational biology?
No full textScientific competitions have driven methodological innovation in artificial intelligence and machine learning, yet remain underutilized in computational biology. This paper provides a comprehensive guide for organizing scientific competitions in bioinformatics, based on our experience with HADACA3, a data challenge focused on deconvolution algorithms for predicting cellular composition in cancer, from multi-omics data. We detail the complete organizational workflow from preparation to execution to postanalysis, offering practical strategies and lessons learned. Beyond advancing multiomics deconvolution methods, HADACA3 demonstrated how expert-guided competitions foster methodological development, scientific exploration, and collaborative networking. By sharing our organizational framework, we aim to democratize scientific competitions as a fundamental methodology for developing and benchmarking computational methods in life sciences.</div
Formation dynamics of an ethylene carbonate-derived solid–electrolyte-interphase in commercial Li-ion batteries
No full textInternational audienceThe importance of the solid–electrolyte-interphase (SEI) is well-established in lithium-ion (Li-ion) batteries, but the technical story behind its formation remains incomplete. Current research has largely focused on the nature of the deposited layer, while the formation dynamics, particularly those occurring in the solution phase, remain elusive. Here, by employing operando infrared fiber evanescent wave spectroscopy (IR-FEWS) to conduct real-time monitoring of the chemical dynamics of ethylene carbonate-based electrolytes and graphite anodes, we reveal that the assembly of the SEI layer follows a classical heterogeneous nucleation and growth process under appropriate kinetic constraints. Our findings, supported by various other in situ/ex situ techniques, show that during charging, the newly generated species (e.g. lithium ethylene dicarbonate (LEDC) and Li2CO3), that are destined for the SEI, can also diffuse away from the graphite–electrolyte interface into the electrolyte. The deposition of the species occurs via a heterogeneous nucleation process with the low-solubility inorganic species (e.g. Li2CO3) preferentially nucleating on the graphite surface, followed by more-soluble organic species (e.g. LEDC). Limiting diffusion to promote the deposition is crucial for facilitating efficient SEI formation with competitive deposition kinetics depending not only on the charging rate and temperature, but also the electrolyte quantity. When the formation parameter-space is intentionally modified by employing a high current pulse during initial charging followed immediately by an ageing step, a more stable SEI with lower resistance is developed, leading to longer lifetimes for the Li-ion pouch cells prepared with this new protocol. Collectively, these findings deepen our mechanistic understanding of SEI formation from the “solution” phase perspective and offer an enriched framework for defining initial charging protocols for battery manufacturing
Antioxidant, Antidiabetic and Topical Antiinflammatory Effects of Hibiscus sabdariffa L. Extracts
No full textInternational audienceBackground:The purpose of this study was to evaluate the antioxidant and antidiabetic activities of an ethanol extract of Hibiscus sabdariffa L. (EEHS) and to compare its topical anti-inflammatory activity to a cream formulated from the plant’s essential oil (EO).Methods:Dried calices powder of the plant was subjected to ethanol extraction using a Soxhlet extraction apparatus. Total phenolic content and total anthocyanin content were assessed. The antioxidant activity was evaluated through a DPPH free radical scavenging assay. EEHS was studied for antidiabetic activity in streptozotocin induced diabetic rats by oral administration of the extract (300 mg/kg body weight) for 15 days. Blood glucose level was estimated and compared with Glucophage (500 mg). In order to understand the biological activity of H. Sabdariffa in the inflammatory response, cream of the EEHS and EO were formulated. Cream was made into oil in water base (O/W) containing 1% and 0.5% (m/m) concentration of EEHS and EO, respectively. An acute dermal irritation test on rabbits of the formulated creams was carried out. The topical anti-inflammatory effect of the creams formulated was determined by the croton oil-induced mouse ear edema test.Results:EEHS containing polyphenolic and anthocyanin compounds exhibited antioxidant properties. The IC50 value of H. sabdariffa was found to be 0.221, indicating its potent inhibitory activity in the tested assay. EEHS at a dose of 300 mg/kg significantly reduced blood glucose levels in hyperglycemic rats from 333 ± 10.226 to 183.33 ± 10.345 mg/dL (P < 0.05 vs. diabetic control) after 14 days of treatment. This reduction was greater than that observed in the positive control group, with glucose reduction percentages of 44.94% and 22.96%, respectively, on day 14. Moreover, EEHS was observed to be non-toxic at doses up to 5 g/kg body weight. In acute inflammation, EEHS demonstrated greater efficacy than EO, achieving inhibition rates of 70% and 66.66%, respectively. The histological study of tissues obtained from mouse ears provided confirmation of the results obtained during this study.Conclusion:The findings demonstrate that EEHS, rich in polyphenols and anthocyanins, exhibits significant antioxidant, antihyperglycemic, and anti-inflammatory properties along with a favorable safety profile - supporting its potential as a natural therapeutic agent
Semi-supervised coronary artery segmentation based on latent information cross-supervision
No full textInternational audienceAccurate and automated segmentation of coronary arteries from coronary computed tomography angiography (CCTA) is essential for diagnosing coronary artery disease (CAD). However, the complex geometry of coronary arteries makes obtaining precise annotations challenging and can reduce segmentation accuracy. This study aims to develop a semi-supervised segmentation method to improve coronary artery segmentation. To achieve this goal, we propose the latent information segmentation network (LIS-Net), a novel 3D semi-supervised framework for precise coronary artery segmentation. LIS-Net integrates frequency-guided blocks to capture latent low-frequency structural features while preserving high-frequency detailed information in CCTA images. This innovation compensates for the limitations of traditional convolutional neural networks (CNNs), which primarily focus on local features in the spatial domain. The framework employs dual decoders specialized for low-frequency and high-frequency representations. The output of each decoder serves as a pseudo-label for the other, establishing cross-supervision that ensures prediction consistency. Additionally, we introduce a multi-scale feature perception (MSFP) module to enhance the encoder's ability to capture global features and propose a coronary morphology loss (CMLoss) function to ensure continuous segmentation. On public datasets ASOCA and ImageCAS, LIS-Net achieved Dice scores of 79.53% and 76.59%, respectively, and 80.82% on an in-house dataset. Compared with existing semi-supervised methods, it demonstrates better overall performance. To our knowledge, this is the first semi-supervised method integrating wavelet-based feature extraction with CNNs for coronary artery segmentation. Evaluations across multiple datasets confirm the effectiveness of LIS-Net and suggest its potential to support more accurate CAD diagnosis in clinical practice. Code is available at https://github.com/YueZhang1029/LIS-Net
Assessing the Limitations of Activation Clipping for Fault Mitigation in Vision and Language Transformers
No full textInternational audienceDeep Neural Network (DNN) activation clipping is a well-established method for mitigating hardware-induced faults during inference. Clipping constrains activations to predefined ranges and filtering corrupted values, including NaN and Inf. However, high variability in activation distributions, both across and within layers of a model, makes it challenging to define fixed ranges that reliably capture corrupted values beyond extreme cases. In this study, we show that activation clipping alone is insufficient to protect vision and language Transformers from inference faults, and that inputs exhibit markedly different levels of fault sensitivity depending on the model's confidence. Our experiments show that the missclassification rate can reach 10.28% even with per-layer clipping is employed
EMG-to-torque models for exoskeleton assistance: a framework for the evaluation of in situ calibration
No full textInternational audienceIn the field of robotic exoskeleton control, it is critical to accurately predict the intention of the user. While surface electromyography (EMG) holds the potential for such precision, current limitations arise from the absence of robust EMG-to-torque model calibration procedures and a universally accepted model. This paper introduces a practical framework for calibrating and evaluating upper-limb EMG-to-torque models, accompanied by a novel nonlinear model. The framework includes an in situ procedure that involves generating calibration trajectories and subsequently evaluating them using standardized criteria. A comprehensive assessment on a dataset with 17 participants, encompassing single-joint and multi-joint conditions, suggests that the novel model outperforms the others in terms of accuracy while conserving computational efficiency. This contribution introduces an efficient model and establishes a versatile framework for EMG-to-torque model calibration and evaluation, complemented by a dataset made available. This further lays the groundwork for future advancements in EMG-based exoskeleton control and human intent detection
Who are the healthcare professionals involved in interprofessional team meetings in French multidisciplinary primary care centres? A quantitative analysis of eight centres
No full textInternational audienceBACKGROUND: Interprofessional collaboration is crucial for providing high-quality care to patients with complex conditions in primary care. In France, multidisciplinary primary care centres (MPCC) receive funding if they organise at least 6 interprofessional team meetings (ITM) per year to discuss complex patient situations and collectively define care strategies. It remains unclear how ITM have been implemented in France. This study analyses healthcare professionals’ involvement in ITMs within MPCCs. METHOD: A multicenter retrospective quantitative study in 8 French MPCCs based on the analysis of ITM reports for the period from 2018 to 2019 was conducted. RESULTS: 1733 patients’situations (n = 1733 cases) discussed during ITMs were analysed. The 8 MPCCs were heterogeneous in terms of geographical location, creation date, and size (3 MPCCs with > 20,000 patients followed by more than 60 professionals and 3 MPCCs with < 20 professionals). On average, five healthcare professionals attended each ITM, with huge variations among MPCCs. The nurse-general practitioner (GP) pair was central. At least one GP was present at 89% (n = 1469) and nurses at 38.15% (n = 630) of ITMs. Participation of other professionals was less frequent and varied according to the MPCC. Physiotherapists were present at 7.8% of ITMs, and other healthcare professionals were present at < 6% of ITMs, including pharmacists (3% of ITMs). In some cases, healthcare professionals external to MPCC, particularly those related to mental health, were also involved in ITMs. CONCLUSIONS: Depending on the MPCC, the professionals involved in ITM vary widely, with the nurse-GP pair at the centre. The relative absence of certain professionals needs to be analysed in order to encourage interprofessional working. TRIAL REGISTRATION: Not applicable
An intense peak of paraglacial dismantlement of mountain slopes: Insights from dating and volume quantification of rock-slope failure deposits in the Icelandic Westfjords (Dýrafjörður and Önundarfjörður)
No full textInternational audienceParaglacial rock slope failures (RSFs) are prominent processes of landscape evolution in deglaciated terrains, such as the Westfjords of Iceland. This study aims to provide chronological and volumetric data on RSF deposits in the Dýrafjörður and Önundarfjörður fjords, in order to document the magnitude, duration, and geomorphic impact of the intense peak of Early and Middle Holocene paraglacial denudation. By refining the timing of a paraglacial signal, this work contributes to a better understanding of sedimentary production and landscape evolution during the Holocene.A total of 17 RSFs was studied, described and mapped using the Schmidt-hammer exposure-age dating method, calibrated with radiocarbon dating. Surficial block morphometry and volumetric estimates of RSF deposits were derived from field measurements, orthophotography, and high-resolution digital elevation models.RSF ages are concentrated in the Early to Middle Holocene. The vast majority of this activity occurred between 12 and 6 cal. ka BP. During this 6000-year interval, ~83 million m3 of debris were deposited, which accounts for approximately 90% of the total volume (~92.5 M m3) from all 17 RSF sites. This indicates a primary paraglacial adjustment phase characterized by high sediment delivery efficiency. Slope reactivations occurred over periods up to 3400 years, with superimpositions of deposits: these are multi-phased RSFs.Finally, a significant lag of approximately 3000 years is observed between the deglaciation (~10.2 cal. ka BP) and the peak in rock-slope failure activity (8–6 cal. ka BP), which coincides with the Holocene Thermal Maximum climax in Iceland (8.6–5.2 cal. ka BP). The subsequent cessation of major RSFs activity after ~4 cal. ka BP marks the transition to a stable, non-glacial equilibrium
The notion of « Conservative destruction »: cyclical obliteration and definitive closure. New data from Incoronata
No full textInternational audienc
Delegation and Resistance: Intendants and the Centralization of the French State
No full textOver the seventeenth century, France relied mainly on private agents to build its state capacity but it shifted gradually to public agents, the intendants. We document this centralization process. The intendants' appointment occurred at different times across the country. Their alternating arrivals and departures in the généralités created sequences of presence and vacancy that our empirical strategy leverages. Using an original panel dataset, we identify a causal effect of the intendants' presence on tax revenues, tax and food riots. Before the 1635 Edict that permanently installed them, they operated occasional missions. They turned to be effective immediately-increasing tax revenues-but only in the pays d'états, provinces that kept the privilege to negotiate on taxation. Under Richelieu and Mazarin governments, from 1635 to 1660, they resisted as the intendants' presence triggered a decrease in tax revenues. In the less autonomous provinces, the pays d'élections, the intendants' presence led tax revenues to increase but it caused tax riots. Finally, the institution reached its maturity under Colbert and his successors: tax revenues increased sharply everywhere but so did tax riots. Moreover, the regulations taken by the intendants were insufficient to cope with food riots