Lux Research (United States)

Fostamatinib is available in France since October 2021 for the treatment of adult chronic immune thrombocytopenia (ITP). French health authorities requested a 3-year, prospective, multicenter registry to provide real-world evidence about the effectiveness and safety of fostamatinib. Patients' characteristics, treatment response (ongoing exposure to fostamatinib and a platelet count ≥ 30 × 109/L with no rescue in the previous 4 weeks) after 3, 6, 12, and 24 months (M); bleeding; fostamatinib discontinuation; adverse drug reactions (ADRs) and other events of interest have been analyzed. In total, 164 patients were included (median age: 59 years; 55.5% women; 84.1% had previous bleeding; 30 had secondary ITP; 89.0% had chronic ITP). The median ITP duration was 7.2 years and the median number of previous ITP treatments was 6. The response rate was 44.0% (70/159) at M3, 41.9% (62/148) at M6, 32.4% (44/136) at M12 and 20.0% (21/105) at M24. Concomitant treatment (mostlyTPO-RA) was used in > 60.0% of responders at each endpoint. The cumulative discontinuation rate at each endpointwas, respectively, 27.0%, 44.6%, 55.9%, and 76.2%. Seventy-one (43.3%) patients experienced at least one bleeding duringfostamatinib exposure; none was fatal. One hundred adverse drug reactions (8 serious) were observed in 61 (36.7%)patients, including diarrhea in 28 (17.1%) patients, arterial hypertension in 17 (10.4%). Seven thrombosis (4.3%) and40 infections (12 serious) were reported in 25 patients (15.2%), mostly in patients with known risk factors. In conclusion,fostamatinib in combination with TPO-RA should be considered in difficult-to-treat ITP patients. No new safety signal was observed.

This study presents a comprehensive investigation of exact fractional wave solutions and bifurcation analysis for the (3 + 1)-dimensional extended shallow water wave (3D-eSWW) equation with β-derivative, which models nonlinear wave phenomena in fluid dynamics and coastal engineering. Leveraging the flexibility of the fractional derivative, the model provides a more generalized and adaptable framework for describing shallow water wave propagation. The Modified Extended Direct Algebraic Method (MEDAM) is systematically employed to derive a broad spectrum of novel exact analytical solutions. These include the following: dark solitary waves, singular solitons, singular periodic waves, periodic solutions expressed via trigonometric and Jacobi elliptic functions, polynomial solutions, hyperbolic wave patterns, combined dark–singular structures, combined hyperbolic–linear waves, and exponential-type wave profiles. Each solution family is presented with explicit parameter constraints that ensure both mathematical consistency and physical relevance, thereby offering a robust classification of wave regimes under diverse conditions. A thorough bifurcation analysis is conducted on the reduced dynamical system to examine parametric dependence and stability transitions. Critical bifurcation thresholds are identified, and distinct solution branches are mapped in the parameter space spanned by wave numbers, nonlinear coefficients, external forcing, and the fractional order β. The analysis reveals how solution dynamics undergo qualitative transitions—such as the emergence of solitary waves from periodic patterns or the appearance of singular structures—driven by the interplay of nonlinearity, dispersion, and fractional-order effects. These insights are crucial for understanding wave stability, predictability, and the onset of extreme events in shallow water contexts. Graphical representations of selected solutions validate the analytical results and illustrate the influence of β on wave morphology, propagation, and stability. The simulations demonstrate that varying the fractional order can significantly alter wave profiles, highlighting the role of fractional calculus in capturing complex real-world behaviors. This work demonstrates the efficacy of the MEDAM technique in handling high-dimensional fractional nonlinear PDEs and provides a systematic framework for predicting and classifying wave regimes in real-world shallow water environments. The findings not only enrich the solution inventory of the 3D-eSWW equation but also advance the analytical toolkit for studying complex spatio-temporal dynamics in fractional mathematical physics and fluid mechanics. Ultimately, this research contributes to the development of more accurate models for coastal protection, tsunami forecasting, and marine engineering applications.

Background and objective: The rapid expansion of artificial intelligence (AI) in healthcare has resulted in substantial advances in diagnostics, prognostics, clinical decision support systems, and patient monitoring. Despite promising performance, many AI-based systems remain insufficiently understood by clinicians and patients due to their “black-box” nature. This lack of transparency may undermine trust, hinder acceptance, and limit safe integration into routine clinical practice. Explainable artificial intelligence (XAI) has emerged as a response to these challenges by enabling human-interpretable explanations of algorithmic decisions. The objective of this narrative review is to synthesize current evidence on XAI in healthcare, with particular emphasis on its technical foundations, clinical applications, influence on trust and decision-making, and broader social and ethical implications. Scope of review: This review synthesizes literature published between 2019 and 2025 addressing explainability in medical AI. The analysis includes methodological studies, clinical evaluations, human–computer interaction research, and social science investigations related to transparency, trust, accountability, and acceptance of AI systems. Relevant publications were identified through structured searches of PubMed, MEDLINE, Scopus, and Google Scholar using keywords related to explainable AI, interpretability, ethics, trust, and clinical decision support. Findings: XAI methods—including feature attribution, model simplification, counterfactual explanations, and visualization techniques—demonstrate potential to enhance clinician understanding of AI outputs and increase confidence in algorithm-assisted decisions. Evidence suggests that explainability may support diagnostic accuracy, reduce automation bias, and facilitate error detection. However, explainability alone does not ensure trust. Clinical context, user expertise, organizational culture, and regulatory frameworks play critical roles in shaping the adoption and appropriate use of explainable systems. Empirical research addressing patient perspectives remains limited. Conclusions: Explainable AI constitutes an important step toward the responsible and socially acceptable integration of intelligent systems in healthcare. While XAI can enhance transparency and trust, its effectiveness depends on thoughtful design, contextual adaptation to clinical workflows, and alignment with user needs. Further interdisciplinary research is required to standardize explainability approaches, evaluate their real-world impact on clinical outcomes, and address the ethical, legal, and societal challenges associated with medical AI.

Cerebrospinal fluid (CSF) circulation is tightly coupled to cerebral blood flow under the fixed-volume constraint of the cranial vault, with cardiac pulsations and respiration acting as dominant physiological drivers. Disruption of these flow dynamics has been implicated in various neurological disorders, motivating the need for imaging methods that capture CSF and vascular flow simultaneously and in real time. Conventional phase contrast MRI (PC-MRI) provides quantitative CSF velocity measurements but relies on cardiac gating and velocity encoding, which limit temporal resolution, dynamic range, and sensitivity to non-cardiac fluctuations. Here, we introduce self-gated ultrafast real-time flow-weighted echo-planar imaging (SURF-EPI), an inflow-weighted approach in which signal intensity reflects the replacement of saturated spins by freshly inflowing, unsaturated water molecules, yielding higher signal with higher flow. This allows simultaneous assessment of arterial inflow, venous outflow, and CSF motion at high temporal resolution. SURF-EPI was acquired at the C2–C3 spinal level, enabling real-time imaging of CSF flow dynamics in the cervical spinal canal alongside arterial and venous blood flow in major cervical vessels (frame-rate: 21.7 Hz). In addition to frequency-domain analyses, we leveraged intrinsic arterial signal fluctuations as a timing reference to reconstruct cardiac-resolved CSF dynamics for individual cardiac cycles without external physiological recordings. Frequency-domain analysis revealed distinct spectral signatures in CSF flow compared with neurovascular flow, including broadened cardiac peaks and enhanced respiratory modulation, particularly within the ventral spinal canal. In contrast, dorsal CSF showed increased power within the cardiac frequency band, higher coherence with cervical vasculature at the cardiac frequency, and reduced non-cardiac contributions. Time-domain analysis showed strong correlation between CSF flow waveforms derived from SURF-EPI and PC-MRI. Beyond ensemble-averaged waveforms, SURF-EPI enabled beat-to-beat analysis, revealing substantial cycle-to-cycle variability in CSF flow that is not captured by time-averaged gated approaches. Together, these findings establish SURF-EPI as a rapid, complementary framework to PC-MRI, enabling real-time neurofluid imaging with integrated time- and frequency-domain characterization of CSF and neurovascular flow dynamics.

Background and objective: Acute respiratory distress syndrome (ARDS) remains one of the leading causes of mortality among patients treated in intensive care units. Despite the use of lung-protective ventilation, prone positioning, and rescue therapies, mortality in severe cases continues to exceed 30–40%. In recent years, the precise optimization of ventilator settings has gained particular importance, including the control of driving pressure, mechanical power, and patient–ventilator asynchrony, while simultaneously the use of advanced monitoring systems and artificial intelligence (AI)–based decision-support tools have expanded rapidly. The aim of this review is to present contemporary, advanced methods of ARDS management, ranging from complex mechanical ventilation strategies to AI-driven solutions, with particular emphasis on their effectiveness, safety, and potential for clinical application. Scope of review: This review includes literature from 2020–2025 addressing interventional and technological aspects of ARDS treatment, with a specific focus on lung-protective ventilation and mechanical power regulation, the application of prone positioning, strategies for reducing patient–ventilator asynchrony, the use of ECMO as a rescue therapy, and AI models designed for outcome prediction, assessment of readiness for ventilator liberation, and automated asynchrony detection. The analysis primarily incorporates observational and cohort studies, methodological work, and publications dedicated to integrating AI into intensive care unit practice and ventilator systems. Findings: Modern mechanical ventilation strategies concentrate on minimizing ventilator-induced lung injury (VILI) by reducing tidal volumes, limiting driving pressure, and decreasing mechanical power. At the same time, accumulating evidence suggests that clinically subtle patient–ventilator asynchrony may worsen outcomes, creating opportunities for machine-learning algorithms capable of its automatic identification. AI models are also being applied to predict weaning success, assess the risk of extubation failure, and develop decision-support systems that integrate ventilator parameters, vital signs, and laboratory data. Although preliminary findings are promising, most AI algorithms remain in developmental stages and require further validation. Conclusions: Current therapeutic approaches in ARDS are shifting away from the exclusive use of lung-protective ventilation principles toward highly personalized strategies incorporating mechanical power assessment, asynchrony analysis, and AI-based predictive tools. The synergistic integration of advanced ventilatory support with AI systems may enable more precise tailoring of therapy to ARDS phenotypes, improve decision-making regarding ventilation management and liberation, and potentially enhance clinical outcomes, however, large, well-designed prospective and randomized studies are needed to determine the impact of these innovations on mortality, ventilator-free days, and the safety of applied interventions.

This study presents a comprehensive simulation-based analysis of optical pilot tone generation for local oscillator synchronization in continuous-variable quantum key distribution (CV-QKD) systems using Gaussian-modulated (GM) coherent states. We compare system performance when the pilot tone is generated either electrically or optically, examining the effects of key parameters such as pilot tone power, laser linewidth, digital-to-analog converter resolution, and link distance. The results provide insights into pilot tone generation strategies that enhance robustness while reducing the implementation complexity of GM CV-QKD systems.

How can a cell navigate its environment without any external cues? Since such cues are not always present in the environment, cells rely on internal machinery to explore their surroundings. Although Rho GTPases are known for orchestrating cell motility, the intrinsic Rho GTPase-effector mechanisms governing spontaneous migration remain incompletely understood. Here we show an imaging-based method that profiles protein-protein interactions (PPIs) through phase-separated condensates. By applying this method to hundreds of interaction profiles between Rho small GTPases and their effector proteins, we uncovered two intrinsic mechanisms governing cell migration. Formin-like protein (FMNL) determines the front of the cell by restricting Cdc42 activity, establishing front-rear polarity. In contrast, Rac1-ROCK-interaction-mediated arc stress fiber formation at the front inherently enables spontaneous directional changes and enhances cellular responses to external cues. Our findings elucidate the intricate roles of the Rho GTPase-effector ensemble that governs cell migration behavior, revealing an intrinsic program for efficient motility strategies.

Digital twins have emerged as valuable tools across industries, enabling real-time monitoring, simulation, and opti-mization of complex industrial systems. Recent developments in generative AI have unlocked significant capabilities, improving the adoption and intelligence of digital twin technology. This paper presents a framework that integrates neural networks, autoencoders, and large language models with aviation digital twin architectures to enable data-centric simulations and machine learning-based prediction capabilities. I described how generative AI can create high-fidelity virtual environments, intelligently impute missing data, simulate unobserved scenarios, and enable zero-shot adaptation across aviation manufacturing and logistics domains. The proposed approach addresses real-world aviation scenarios, including predictive maintenance in aircraft manufac-turing, logistics inventory optimization, and network anomaly de-tection. The framework dynamically integrates multimodal data from Industrial Internet of Things (IIoT) sensors, enterprise re-source planning systems, and supply chain networks. I discussed implementation challenges, ethical and security considerations, scalability issues, and human-centric design requirements for generative AI digital twins in aerospace manufacturing. This research contributes to the conceptual development of next-generation self-aware, self-improving, intelligent cyber-physical systems in aerospace.