Locating and identifying underwater bodies is challenging because underwater images are often low-luminance, turbid, and have ambiguous backgrounds. Object detection and computer vision technologies are gaining popularity in marine biology for the quick and efficient detection, and processing in the aquatic environment, despite these drawbacks. To overcome these limitations, a novel hybrid framework, Quantum (Conceptual) Dilated Convolutional Kronecker Networks (QDCKN) is proposed that integrates quantum-inspired dilated convolutions, fuzzy uncertainty modeling, and Kronecker-structured feature compression in a mathematically coupled architecture. The proposed framework follows a integrated two-stage underwater object analysis framework consisting of Savitzky-Golay filtering for image pre-processing and YOLOv3 for object detection followed by a QDCNN-Fuzzy Deep Kronecker Network (DKN) module for region-wise classification. Unlike conventional approaches that treat feature extraction and classification independently, the proposed method establishes a cohesive interaction between quantum-inspired feature representation, fuzzy logic-based uncertainty handling, and structured feature compression, enabling robust recognition under non-uniform illumination, occlusions, and noise. An experimental evaluation conducted within the evaluated dataset (Underwater Object Detection dataset) demonstrates that QDCKN achieves a maximum classification accuracy of 93.40%, precision of 92.20%, and recall of 94.60%, outperforming SA-FPN, MarineDet, mResNet, and EDR. The experimental analysis and ablation study proves that the proposed work contributes to improved performance and stability under challenging underwater conditions within the evaluated dataset.

Deep Learning in Dental Imaging: Advances, Challenges, and Future.

Artificial intelligence and data-driven models are changing hepatology, but expert clinical judgment remains essential. Liver diseases are complex and evolve over time, requiring reasoning that links numerical data with clinical context. We propose data-informed intuition (DII), a conceptual framework bridging evidence-based medicine with clinical experience and reflective reasoning. Instead of presenting a validated prediction model, it proposes DII as an integrative scaffold for clinical reasoning that can be examined and revised. DII is viewed as a trainable skill that improves with data feedback, not mere guesswork. Clinicians continually calibrate their judgment as patient information accumulates. In hepatology, DII helps interpret disease trajectories, adapt predictive scores to context, and handle uncertainty in liver transplantation, hepatocellular carcinoma, acute liver failure, and related conditions. The DII framework describes eight axes (time, rate, magnitude, pattern, context, causality, integration, and uncertainty) that make visible how data and intuition interact in clinical reasoning. Integrating DII into fellowship training and artificial intelligence-assisted systems could nurture clinicians who combine algorithmic precision with human insight. This perspective summarizes the conceptual basis, clinical uses, and educational value of DII, and outlines both a research agenda for its empirical testing and a path toward data-informed professionalism in hepatology.

Interoceptive machine framework: Toward interoception-inspired regulatory architectures in artificial intelligence.

Advances in automated image classification, together with near-global imaging coverage of the Martian surface, have enabled detailed characterization of the spatial distribution of pitted cones, providing a foundation for systematic investigations of their morphological diversity. Concurrently, the continued acquisition of high-resolution Martian imagery over the past decades has allowed photogrammetrically derived digital elevation models (DEMs) to enhance the accessibility, precision, and overall robustness of morphological analyses. However, the number of identified Martian pitted cones causes systematic, manual morphological measurements to be highly labour-intensive and, consequently, impractical for large datasets. To address this challenge, we present a command-line open-source MarsCONE software toolbox, designed to perform automatic cone-morphology detection, and to compute the morphological parameters of Martian pitted cones using High Resolution Imaging Science Experiment (HiRISE)-derived DEMs. The toolbox is built on as a suite of Python tools and Jupyter notebooks, and performs key processing steps including data preparation and transect generation according to user-defined configurations (Generator), signal extraction and landform-point morphology detection (Finder), and cross-transect aggregation with uncertainty handling and data export (Analyzer). This enables MarsCONE to analyze hundreds of pitted cones in a single batch, providing fully reproducible and systematic results within seconds and at minimal computational cost. Consequently, the MarsCONE toolbox improves reproducibility, reduces manual workload, and facilitates large-scale comparative studies of pitted cones across Mars, thereby supporting a more robust understanding of the geological processes governing their formation.

ABSTRACT Crisis communication is often examined as a reputational, informational, or public‐relations function. In safety‐critical and crisis‐prone systems, however, some communication events may also perform safety‐control functions by supporting detection, warning, coordination, public protection, mitigation, recovery, accountability, and learning. This article develops an exploratory safety‐barrier framework for analysing crisis communication across organisational crisis cases. Drawing on a comparative event‐sequence dataset of 60 coded communication events across six cases—Deepwater Horizon, Boeing 737 MAX, Fukushima Daiichi, Grenfell Tower, East Palestine, and BP Texas City—the study codes communication actors, crisis phases, message types, safety‐control functions, communication quality, uncertainty handling, relational dependencies, barrier effects, and escalation effects. The analysis identifies three communication‐barrier states: protective communication, degraded communication, and pathologic communication. Protective communication supports safety control through timely, transparent, coordinated, uncertainty‐aware, and actionable messaging. Degraded communication provides partial or delayed support and leaves safety‐control functions incomplete. Pathologic communication actively contributes to crisis escalation through contradiction, false reassurance, defensive framing, distorted feedback, or suppressed dissent. The article contributes to crisis communication and safety studies by proposing communication as an event‐level safety‐control function embedded in interdependent actor networks. It also offers a Communication Barrier Review Matrix for crisis preparation, real‐time coordination, post‐crisis investigation, and safety‐management review.

Artificial intelligence (AI) is increasingly applied to biomedical research, but most current systems remain limited to specific tasks, data types, or biological scales. This makes it difficult to connect molecular alterations, organelle dysfunction, cellular behavior, tissue remodeling, organ physiology, systemic regulation, and whole-body phenotypes into coherent biological reasoning. In this Perspective, we propose the Full-Body AI Agent as a hypothetical multi-agent framework and conceptual blueprint for future systemic biology and precision medicine, rather than a fully implemented software platform. This framework envisions a supervisory Full-Body AI Agent coordinating seven biological-level agents, namely Molecule, Organelle, Cell, Tissue, Organ, Organ System, and Body System AI Agents, to standardize biomedical data, decompose cross-scale questions, assign level-specific tasks, and integrate outputs through iterative feedback. We further outline the data commons, harmonization mechanisms, uncertainty handling, arbitration strategies, and traceability safeguards required for biologically grounded cross-scale reasoning. Two hypothetical scenarios, metastasis analysis and drug development, illustrate how this framework could organize multilevel evidence from molecular changes to systemic phenotypes and therapeutic responses. This Perspective aims to clarify the conceptual basis of full-body AI and provide a foundation for transparent, physiology-constrained, cross-scale AI systems in disease analysis, therapeutic evaluation, and personalized medicine.

Digital twins are increasingly promoted in neurology as an advancement beyond conventional artificial intelligence, yet the term is often applied without conceptual or methodological rigor. Strict definitions describe digital twins as dynamically updated, bidirectionally linked models that generate predictive, decision-relevant value, criteria rarely met by current neurological applications. This Opinion critically examines the state of digital twins across major neurological domains, including dementia, multiple sclerosis, Parkinson's disease, epilepsy, stroke, pain, and migraine. We argue that most existing systems are more accurately described as twin-inspired longitudinal decision-support or trial-analytics models rather than true clinical digital twins. While neurology is well-suited to digital twin approaches due to disease heterogeneity, multimodal data, and iterative care pathways, progress is limited by gaps in measurement validity, uncertainty handling, prospective evaluation, and governance. A pragmatic path forward is proposed, emphasizing question-specific, validated neurological digital twins over overgeneralized brain twin narratives, and suggesting that much of the current field is better understood as twin-inspired modeling rather than true clinical digital twin implementation.

From mathematical to AI-based methods: A review of marine PNT data fusion and uncertainty handling

We develop a new formalism for the component separation method Spectral Matching Independent Component Analysis (SMICA) in order to include the information contained in the foregrounds beyond second-order statistics. We also develop a binned bispectrum estimator that works directly using maps of different frequency channels, capable of determining the bispectrum of multiple components at the same time, shifting the traditional approach to non-Gaussianity estimation from a cleaned map to the component separation step, for a better handling of foreground uncertainty. We test our method on 400 E and B polarization simulations based on the LiteBIRD experiment, containing the two main sources of contamination for CMB polarization experiments: polarized dust and synchrotron emission. We show that the bispectrum does not improve the precision of the power spectrum estimation or of the spectral parameters. However, we are capable of recovering the correct 3-point correlator of the foregrounds and standard constraints on primordial non-Gaussianity in a coherent multi-frequency and multi-component framework. The advantage of our approach is that it combines data in an optimal way accounting for the power spectrum and the bispectrum of the various components, which is not true for the standard approach.

Learning expressive representations for complex, size-varied molecular graphs remains a fundamental challenge in toxic molecular property prediction and regression. The inherent nonlinearity of atomic interactions, together with intricate structural dependencies and latent high-order structural characteristics, makes accurate graph-based learning particularly difficult. Although recent deep learning (DL)/graph neural network (GNN) approaches have leveraged message passing, attention mechanisms, and graph transformer architectures, these techniques still suffer from limited nonlinear expressiveness, insufficient modeling of high-order structural information, and a lack of explicit uncertainty handling. To address these limitations, we propose a novel AKAGTL model, which is an attention-driven Kolmogorov-Arnold network (KAN)-based graph transformer framework for toxic molecular graph embedding and regression learning. Unlike existing approaches that rely on linear or shallow nonlinear transformations, our proposed AKAGTL model introduces a structured KAN-based functional transformation to explicitly model complex high-order atomic interactions within an attention-driven graph transformer backbone. In addition, high-order structural representations are systematically incorporated to complement structural encoding, while a Gaussian neuro-fuzzy fusion mechanism enables uncertainty-aware aggregation of heterogeneous feature spaces. The proposed framework is evaluated on multiple molecular toxicity benchmarks under a unified experimental protocol with repeated runs. Comprehensive experiments within benchmark molecular graph datasets demonstrate that our AKAGTL model can consistently improve regression accuracy compared to representative GNN and graph transformer baselines. These findings suggest that jointly modeling nonlinear functional interactions, structural dependencies, and uncertainty-aware fusion provides a more expressive and robust solution for toxic molecular graph embedding and regression learning.

The hazard identification of chemicals is a key step of the 'Safe and Sustainable by Design' (SSbD) framework introduced by the European Commission, aiming to eliminate hazardous substances early in innovation. In this context, in silico methods such as (Quantitative) Structure-Activity Relationship ((Q)SAR) models offer rapid, cost-effective, and animal-free alternatives for early-stage hazard screening. The Partnership for the Assessment of Risks from Chemicals (PARC) is developing a toolbox to facilitate SSbD assessments containing numerous (Q)SAR models. Challenges, however, exist in using and combining multiple in silico tools. Here, we developed a workflow to assess chemical hazards using multiple in silico tools within the PARC toolbox. The workflow consists of three phases: 1) the preparation stage, 2) running the models, and 3) the evaluation stage. To demonstrate the approach, we applied it to a case study comparing Bisphenol A, Isosorbide, and Bisphenol AP. Tools from the PARC toolbox were screened for relevance, transparency, and open access availability. Only models aligned with SSbD required endpoints and adequately documented via (Q)SAR Model Reporting Formats were retained. The properties assessed in this study cover carcinogenicity, germ cell mutagenicity, reproductive toxicity, endocrine disruption, persistence, bioaccumulation, and aquatic toxicity. Predictions were filtered using applicability domain criteria and reliability scores. Next, three strategies were applied for integrating different model outputs. Model agreement varied across endpoints and integration methods. This emphasizes the possibility of different SSbD assessment outcomes and thus the need for transparent documentation of the chosen strategy and explicit handling of uncertainty. Our study demonstrates how multiple models can systematically and transparently be integrated via the developed workflow. Key areas for improvement are to refine integration strategies, harmonize the definition and communication of applicability domains across tools, expand in silico coverage for currently underrepresented endpoints, and to develop approaches to consider data gaps in SSbD assessments.

ABSTRACT Maintenance and reliability of mining equipment are critical for meeting production targets, managing costs, and supporting safe operations. This paper provides a structured review of maintenance and reliability research in mining, focusing on single-unit (SU) and multi-unit (MU) modelling approaches. Using this distinction as an analytical framework, 87 peer-reviewed studies published between 2000 and 2024 are analysed in terms of objectives, solution methods, and consideration of workforce and spare-parts constraints. The review shows a shift towards data-driven and simulation-based methods and identifies gaps in decision-oriented modelling, uncertainty handling, and coordination with production planning.

Abstract – Multi-Agent Systems (MAS) have emerged as a powerful and flexible paradigm within distributed artificial intelligence, enabling the efficient solution of complex, large-scale, and dynamic problems by decomposing them into smaller, manageable subtasks handled by multiple autonomous agents. An agent can be defined as an independent computational entity capable of perceiving its environment through available inputs, processing this information using predefined rules or learning algorithms, and taking appropriate actions to achieve its assigned objectives. In a MAS, these agents do not operate in isolation; instead, they interact, collaborate, and sometimes compete with one another, sharing knowledge and coordinating their actions to achieve both individual and collective goals. This collaborative behavior significantly enhances system performance, allowing MAS to exhibit key characteristics such as scalability, adaptability, robustness, and fault tolerance, which are essential in real-world applications. The distributed nature of MAS reduces the dependency on a central controller, thereby eliminating single points of failure and enabling the system to continue functioning even if some agents fail or behave unexpectedly. MAS have been widely applied across diverse domains including computer networks, cloud computing, robotics, smart grids, transportation systems, social networks, and urban infrastructure, where they facilitate intelligent decision-making, efficient resource management, and real-time problem-solving. For instance, in cloud computing, agents can dynamically allocate resources and balance workloads, while in robotics, they enable coordinated movement and task execution among multiple robots. Furthermore, the integration of advanced learning techniques such as reinforcement learning and evolutionary algorithms allows agents to adapt to changing environments, improve their decision-making over time, and handle uncertainty more effectively. However, despite their numerous advantages, MAS also face several significant challenges that must be addressed to fully realize their potential. These include coordination and consensus among agents, efficient communication in large-scale systems, task allocation based on agent capabilities, fault detection and isolation, maintaining system security, and managing dynamic and unpredictable environments. Additionally, issues such as scalability, synchronization, and maintaining connectivity among agents further complicate system design and implementation. This comprehensive exploration of Multi-Agent Systems provides an in-depth understanding of their fundamental principles, architectural characteristics, and operational mechanisms, along with a detailed examination of their applications and associated challenges. By analyzing both theoretical foundations and practical implementations, this work offers valuable insights into the design and development of intelligent, distributed systems, making it a useful resource for researchers, engineers, and practitioners aiming to build advanced agent-based solutions in modern computing environments. Key Words: Multi-Agent Systems (MAS), Autonomous Agents, Distributed Artificial Intelligence, Agent Communication, Coordination and Collaboration, Task Allocation, Scalability, Adaptability, Reinforcement Learning, Cloud Computing, Robotics, Smart Systems, Fault Tolerance, Security, Intelligent Decision Making.

Clustering ensemble improves clustering quality by integrating multiple base clustering results; however, existing methods suffer from inadequate handling of boundary uncertainty and lack a unified probabilistic-to-decision framework. This paper proposes GMM-3WD-CE, which integrates Gaussian Mixture Model (GMM) with three-way decision (3WD) theory to construct a multi-level uncertainty modelling framework. The method generates [Formula: see text] diverse base clusterings via a multi-algorithm strategy, constructs a weighted co-association matrix using quality scores derived from the silhouette coefficient, the Caliński-Harabasz index, and the Davies-Bouldin index, employs the ICL criterion for optimal GMM model selection, and adaptively calculates three-way decision thresholds through the Otsu algorithm to partition samples into core, boundary, and trivial domains. Differentiated label-assignment strategies for each region yield the final consensus clustering. Comparative experiments on eight benchmark datasets with nine comparison methods show that GMM-3WD-CE achieves statistically significant average improvements of [Formula: see text] in NMI and [Formula: see text] in ARI over PCPA and [Formula: see text] in NMI and [Formula: see text] in ARI over classical MCLA, while remaining competitive with the strongest recent baseline, SDGCA ([Formula: see text] average NMI advantage; Wilcoxon [Formula: see text], medium effect size [Formula: see text]). Ablation experiments verify the contribution of each component; Wilcoxon and Friedman tests with Cohen's d effect sizes confirm statistical significance against all other baselines; and runtime/scalability analyses characterise the computational trade-offs.

Control barrier functions (CBFs) provide a rigorous framework for enforcing safety in control-affine systems by ensuring system states remain within predefined safe sets. However, practical deployment faces fundamental challenges that limit real-world applicability. This review analyzes recent progress in CBF methodologies across three interconnected domains: uncertainty handling, structural optimization, and feasibility assurance. For uncertainty, we distinguish strategies tailored to unknown dynamics, modeling discrepancies, and dynamic environments, spanning robust theoretical methods and learning-based approaches. For structural design, we examine class- $\mathcal {K}$ function selection, parameter tuning, and advanced modifications that jointly address conservatism and feasibility. For feasibility, we identify the root causes of CBF-QP infeasibility and survey solution strategies, including constraint relaxation, structural redesign, and mathematical guarantees. By synthesizing these directions into a unified framework, this review highlights key interdependencies and outlines future research opportunities for advancing CBF-based safety-critical control in robotics, autonomous systems, and beyond.

Modeling Approaches to the Economic Evaluation of Microelimination Strategies for Hepatitis C Virus in High-Risk Populations.

Energy management strategy (EMS) has a significant impact on the lithium-ion battery/supercapacitor combined hybrid energy storage systems (HESS) of electric vehicles. However, existing strategies often suffer from inefficient energy allocation and poor dynamic response. To address these challenges, this paper proposes a fuzzy reinforcement learning (RL)-based EMS to optimize the efficiency of HESS. Firstly, key road features are extracted from vehicle operation data using a sliding window, and then principal component analysis and an optimized K-nearest neighbors’ algorithm are used for feature dimensionality reduction and clustering, which can enable more precise feature classification under various road conditions. Secondly, a dueling double deep fuzzy Q-network (D3FQN) framework is proposed for real-time power allocation optimization. By combining the uncertainty handling capability of the fuzzy inference system with the efficient learning ability of the dueling architecture, the real-time adaptively decision-making ability of HESS can be enhanced under complex road conditions. Additionally, an adaptive reward function is designed to dynamically adjust the allocation strategy according to real-time road conditions, power demand, and battery status. Finally, the physical and comparative experiments are conducted, and the results validated the strong dynamic adaptability and efficient energy allocation of the proposed strategy across various road conditions.

This study introduces a Lean Multi-Agent Deep Reinforcement Learning (L-MADRL) framework for energy management of networked microgrids (NMGs) with multiple electricity retailers (ERs) and microgrids (MGs) under renewable and load uncertainties. Coordinating energy exchanges in such systems is challenging due to the need for market efficiency, technical feasibility, and scalability. The proposed framework combines a multi-agent Deep Q-Network (DQN) with a single-level reformulation of a bi-level optimization model. In this formulation, the upper level maximizes ER profits and the network’s available transfer capability (ATC), while the lower-level MG cost minimization is replaced by Karush–Kuhn–Tucker (KKT) conditions, yielding a mathematical program with equilibrium constraints (MPEC). This hybrid design offers two benefits: (i) technical constraints such as power flow limits, generator capacities, and market rules are embedded in the MPEC, freeing DRL agents from constraint enforcement and improving learning stability and policy reliability, and (ii) explicit ATC consideration enhances power transfer efficiency and enables network-aware coordination. Performance was evaluated on PJM 5-bus and IEEE 14-bus test systems against deterministic, risk-neutral (RNSO), and risk-averse (RASO) stochastic optimization. Results show that in the 5-bus case, L-MADRL reduced MG costs by 10.3% and increased ER profits by 3.7%, while in the 14-bus case costs decreased by 2.6% and profits rose by 11.4%. L-MADRL also improved ATC, exceeding the best benchmark by 32% in the 5-bus system and by 30% initially and 10% at peak in the 14-bus system. Across all cases, runtimes remained below 3 s, highlighting the framework’s scalability and computational efficiency. • Lean multi-agent DRL (L-MADRL) framework developed for networked microgrid energy management. • Integrates DQN with single-level KKT-based reformulation of a bi-level problem. • Explicit inclusion of ATC ensures secure and network-aware power exchanges. • Achieves cost reductions of 10.3% (5-bus) and 2.6% (14-bus) for microgrids. • Increases ER profits by up to 11.4% while maintaining runtimes under 3 s.

Surgical consultations contain dense information that patients struggle to recall, and existing tools do not provide access to individualized clinical dialogue. Retrieval-Augmented Generation (RAG) can ground language-model responses in source transcripts, enabling patients to revisit the specific advice given during their consultation. This study evaluates the technical feasibility and performance of a RAG conversational agent built on synthetic plastic surgery encounters. Twenty simulated patient cases were generated across ten plastic surgery procedures, each containing four sequential visits. Transcripts were converted to audio and transcribed using Whisper to simulate real-world ambient capture. The transcriptions were ingested into a Vertex AI RAG system powered by Gemini 2.0 Pro. Two hundred patient-style questions were developed across domains including procedural details, risks, recovery, medications, and clinical parameters. Responses were evaluated against gold-standard transcript answers for accuracy, precision, recall, and error type. Readability was assessed using standard metrics. The system achieved 99.0% accuracy (198/200), with perfect precision and 0.99 recall. One question demonstrated appropriate uncertainty handling when terminology differed from transcript content. No hallucinations or incorrect substitutions occurred. Readability analysis showed a mean Flesch-Kincaid Grade Level of 8.56 and a Reading Ease score of 60.87, indicating accessible patient-level language. A RAG conversational agent grounded in consultation transcripts can deliver highly accurate, patient-friendly recall support, with strong factual reliability and appropriate uncertainty handling. Temporal integration remains the primary area for improvement. These findings demonstrate the feasibility of transcript-based patient assistance and support future development toward real-world deployment.