搜索结果：AI-Driven

Quadruped robots have attracted increasing attention because they can traverse uneven terrain, support field deployment, and perform tasks that are difficult for wheeled or tracked platforms. Recent advances in artificial intelligence (AI) have further expanded their capabilities from manually designed gait control toward learning-based locomotion, perception-aware adaptation, dynamic motion skills, autonomous recovery, manipulation, energy-aware operation, fault diagnosis, and human-robot interaction. However, the literature on AI-driven quadruped robotics is distributed across diverse technical topics, robot platforms, validation settings, and performance metrics, making it difficult to assess the maturity and practical value of different approaches. To address this need, this review provides an AI-centered and deployment-oriented overview of quadruped robotics. A systematic literature search was conducted using Web of Science, IEEE Xplore, ACM Digital Library, ScienceDirect, and SpringerLink, covering studies published approximately from 2000 to 2025. After screening and eligibility assessment, 287 studies were included for detailed review. The review first examines AI-driven locomotion, including reinforcement learning, non-RL machine-learning methods, model-based approaches, and hybrid strategies, with attention to robustness, sim-to-real transfer, sensor use, computational requirements, and hardware validation. It then summarizes AI-supported advanced behaviors, including jumping, fall prevention and recovery, and object manipulation, focusing on reported quantitative performance, impact management, and reliability. Finally, it discusses system-level topics that affect real-world deployment, including fault diagnosis, energy-efficient control, shared autonomy, trust-aware and explainable interaction, and safety-aware human-robot collaboration. By organizing the literature according to robot capabilities, validation maturity, and deployment challenges, this review helps clarify the current progress, limitations, and future directions of AI-driven quadruped robots.

The escalating global mental health burden has triggered an urgent real-world need for scalable artificial intelligence (AI) solutions. This study employs a quantitative bibliometric approach to objectively present the research hotspots and emerging trends in the global AI-driven mental health landscape (2016-2025), specifically capturing the post-2023 surge in generative AI. We analyzed 9,623 original research articles from Web of Science and Scopus. CiteSpace and VOSviewer were utilized for co-authorship analysis, literature co-citation clustering, and keyword burst detection. Research output grew exponentially, surpassing 3,000 publications in 2025. The US and China led productivity, with King's College London as the top institution. Research clusters encompass foundational AI methodologies and four clinical domains: diagnosis (e.g., neuroimaging), digital interventions (e.g., chatbots), targeted conditions (e.g., PTSD), and suicide prevention. Recent keyword bursts highlight diagnostic accuracy, loneliness, and occupational burnout as post-2023 frontiers. The field is transitioning from exploratory algorithmic development to patient-centered clinical applications. While AI offers transformative potential for mental health research and care, future research must urgently prioritize patient safety, algorithmic bias, data privacy, safeguards against algorithmic hallucinations, and ethical integration into clinical workflows to ensure sustainable implementation. What is already known about this topic: (1) The escalating global mental health crisis urgently requires scalable clinical interventions to overcome severe shortages in traditional healthcare resources and personnel. (2) Artificial intelligence offers a transformative solution to bridge this treatment gap, with applications rapidly expanding across psychiatric diagnosis, treatment, and active interventions. (3) Despite exponential growth in AI applications, the knowledge base remains fragmented; practitioners lack objective, large-scale evidence to guide clinical integration. What this topic adds: (1) By analyzing 9,623 articles (2016–2025), this study systematically maps the field’s structural evolution from predictive algorithms to generative intelligence. (2) The research identifies key clinical hotspots like chatbots and suicide prevention, highlighting diagnostic accuracy, loneliness, and occupational burnout as emerging post-2023 frontiers. (3) By systematically categorizing research into five functional domains, this study provides a clear taxonomy, offering practitioners a valuable reference to understand AI’s practical applications.

Antimicrobial Resistance (AMR) has evolved from a clinically observed phenomenon into a complex, dynamic, and partially predictable evolutionary process. Traditional approaches centered on phenotypic detection and retrospective surveillance are increasingly inadequate to address the accelerating pace of resistance emergence. This review presents a paradigm shift toward predictive antimicrobial science, driven by the convergence of Evolutionary Intelligence (EI), Artificial Intelligence (AI), genomic surveillance, molecular simulation, and digital twin technologies. Leveraging whole-genome sequencing (WGS) and resistome analytics, AI models can identify latent resistance determinants and forecast evolutionary trajectories before clinical manifestation, enabling a transition from reactive to anticipatory intervention strategies. Central to this transformation is the concept of the Computational Antimicrobial Resistance Ecosystem (C-AMRE), an integrated, multi-layered framework that unifies data acquisition, predictive modeling, mechanistic simulation, and clinical feedback into a continuous learning system. Within this ecosystem, molecular simulations provide mechanistic insights into resistance at atomic and systems levels, while AI-driven pharmacology enables the design of novel antibiotics, antimicrobial peptides, and Nano-Adjuvants through generative and optimization-based approaches. The incorporation of digital twins further advances precision medicine by simulating patient-specific infection dynamics, pharmacokinetics/pharmacodynamics (PK-PD), and resistance evolution in real time, thereby enabling adaptive and personalized therapeutic strategies. Across micro-, meso-, and macro-scales, these technologies collectively redefine AMR as a systems-level phenomenon that can be modeled, predicted, and strategically managed. However, challenges related to data integration, model interpretability, validation, ethical governance, and global accessibility remain critical barriers to implementation. Despite these limitations, the integration of AI and computational frameworks positions antimicrobial research at the forefront of a new era, where antibiotics are no longer static interventions but adaptive components of intelligent, continuously evolving systems. This review highlights the transition from detection to prediction and ultimately to adaptive intervention, emphasizing the role of computational ecosystems in shaping the future of sustainable antimicrobial therapy.

Atopic dermatitis (AD) affects around 20% of children and up to 10% of adults. Its fluctuating course, severe pruritus, and impact on sleep, mental health, and daily functioning highlight the need for innovative approaches to diagnosis and disease management. Digital health technologies have rapidly expanded to fill this gap, yet their clinical readiness remains uncertain. This scoping review aims to map AI-driven digital tools for AD, classify their functionalities, assess methodologies, and identify challenges and opportunities for real-world implementation. A comprehensive search of MEDLINE (Ovid), Embase (Ovid), Web of Science, and Scopus was conducted from database inception to December 2024 using controlled vocabulary and free-text terms related to "atopic dermatitis," "eczema," "artificial intelligence," "digital applications," and "digital tools." Two reviewers independently screened studies and extracted data, with conflicts resolved by a senior reviewer. Eligible studies included primary research on diagnostic, symptom-tracking, predictive, teledermatology, or language-based tools. Methodological quality was assessed using a 20-item standardised framework. Descriptive statistics and summary tables were used to synthesise findings. 52 studies met inclusion criteria with some having multiple applications: diagnostic tools (n=32), symptom-tracking tools (n=8), predictive models (n=16), teledermatology tools (n=3), and Natural Language Processing/Large Language Models (NLP/LLM)-based applications (n=2). While most studies reported methodological transparency (96%) and data partitioning for model development and evaluation (96%); external validation (21%), code availability (21%), skin colour reporting (27%) and multi-expert labelling (38%) were limited. Despite several Convolutional Neural Network (CNN) based diagnostic models achieving >90% accuracy, few tools underwent real-world testing or clinical integration. AI-driven tools for AD show strong promise. However, limited validation, insufficient amounts of variation in skin type in source data, and real-world evaluation restrict clinical translation. Collaborative efforts to strengthen methodologies, improve dataset representativeness, and evaluate tools in clinical settings are essential for effective implementation.

To compare the clinical outcomes of vascularized free fibula flap (FFF) and deep circumflex iliac artery (DCIA) flap for mandibular reconstruction, using a novel AI-driven dual-modality system for quantitative 3D morphological fidelity and facial symmetry analysis. We evaluated 56 patients with tumors (34 undergoing FFF and 22 undergoing DCIA) using a 3D U-Net deep learning model and Face Mesh. The U-Net model predicted mandibular shapes to generate patient-specific, anatomically plausible reference mandibles, and surface deviations were compared with the actual reconstructions. Facial symmetry was quantified using MediaPipe Face Mesh. No significant differences in facial symmetry (S&#x2009;=&#x2009;0.084&#x2009;&#xb1;&#x2009;0.013 vs. 0.083&#x2009;&#xb1;&#x2009;0.012, p&#x2009;=&#x2009;0.781), functional recovery, or complications. DCIA showed lower morphological deviations for body defects: mean difference (1.42&#x2009;&#xb1;&#x2009;0.83&#xa0;mm vs. 2.58&#x2009;&#xb1;&#x2009;0.62&#xa0;mm, p&#x2009;=&#x2009;0.001), maximum deviation (19.8&#x2009;&#xb1;&#x2009;2.7&#xa0;mm vs. 21.4&#x2009;&#xb1;&#x2009;4.1&#xa0;mm, p&#x2009;<&#x2009;0.05). FFF and DCIA flaps achieve comparable facial symmetry, functional recovery and complication rates in mandibular reconstruction. DCIA flap offers significantly improved 3D geometric accuracy only for isolated mandibular body defects, while the clinical relevance of this statistical difference remains uncertain. The AI-based quantitative assessment system established in this study provides an objective tool for defect-specific flap selection in mandibular reconstruction.

This study presents an AI-driven framework for multi-class disease detection from unstructured, patient-reported textual symptom descriptions, combining natural language processing (NLP) with five machine learning classifiers: Neural Networks, Decision Trees, Logistic Regression, Multinomial Na&#xef;ve Bayes, and Gradient Boosting. The core strength of the proposed framework lies in its unified, lightweight, and interpretable pipeline that integrates TF-IDF-based symptom text representation with classical and neural machine learning models, introduces a controlled text-generation mechanism using positive and negative templates with noise injection, and enables systematic cross-modal comparison across UMLS-derived, structured categorical, and synthetically generated textual symptom datasets. The framework is evaluated across three complementary dataset representations: (i) a UMLS symptom-disease knowledge base covering 149 diseases and 404 symptoms, (ii) a structured Kaggle categorical dataset comprising 42 disease classes and approximately 5000 samples described by 132 symptom features, and (iii) a synthetically generated text-based symptom dataset derived from the Kaggle data. All models are evaluated using accuracy, precision, recall, and F1-score. On the synthetically generated text-based test set under controlled conditions, the Neural Network achieves the highest performance, with an accuracy of 99.3% and F1-score of 99.3%, followed by Multinomial Na&#xef;ve Bayes with an accuracy of 99.0%. On the Kaggle categorical dataset, Logistic Regression attains the highest accuracy of 98.9%, demonstrating that lightweight and interpretable models can perform competitively on structured symptom representations. It is important to note that the text dataset is synthetically generated and does not comprise authentic unstructured clinical patient narratives; results should therefore be interpreted with appropriate caution regarding real-world generalizability. Although high accuracy values are obtained, these results should be interpreted in light of dataset-level limitations, including class imbalance in the Kaggle categorical dataset, the synthetic nature of the text-based symptom dataset, and the absence of external validation on authentic patient-reported clinical narratives. Overall, the proposed pipeline offers a scalable foundation for telemedicine, mobile health triage, and low-resource clinical settings where patients express symptoms in natural language rather than selecting from predefined lists.

Artificial intelligence (AI) in healthcare is assumed to introduce risks that are not easily addressed by dominant philosophical models for thinking about responsibility. When an AI tool makes an error that results in patient harm, the question of who is responsible is rarely straightforward. Dominant models of responsibility work when harm can be traced to a single actor, but they fail in socio-technical systems where decisions and actions are distributed across multiple human and technological agents. Iris Marion Young's social connection model of responsibility offers a systematic way to rethink responsibility in these contexts. Rather than locating responsibility solely after harms occur, Young proposes a forward-looking account that attaches responsibility to those who participate in and benefit from the structural processes that produce harm. Young's framework organised around parameters of reasoning-power, privilege, interest, collective ability, and personal connection- provides an alternative way for understanding differentiated responsibilities across core actors in AI-driven healthcare.

Intracerebral Haemorrhage (ICH) is one of the most deadly forms of stroke, with a high risk of morbidity and fatality. Due to increased anticoagulant use over the past four decades, the global incidence of hemorrhagic stroke has increased. Chronic hypertension and diseases like cerebral amyloid angiopathy are commonly associated with ICH. PRISMA criteria were adhered to in this review. PubMed, Web of Science, Scopus, and Google Scholar were used to conduct a thorough literature search. Neural networks, artificial intelligence, diagnostics, machine learning, and intracerebral haemorrhage were among the keywords utilized. Only English-language, peer-reviewed studies published after 2018 were included. After title, abstract, and full-text screening, research focused on AI-based ICH diagnosis and treatment was selected, whereas studies unrelated to or subpar were excluded. Studies on AI-assisted diagnosis, machine learning applications, phytocompound-based therapies, and novel therapeutic approaches for intracerebral haemorrhage were compiled. 32 synthetic compounds, due to their multiple modes of action, including lowering Oxidative stress, minimizing swelling, improving hematoma drainage, and preserving the blood-brain barrier, are promising therapeutic agents. Prognostic biomarkers, ultra-early hemostatic therapy, perihematomal protection against inflammatory brain injury, minimal invasive surgery, and primary prevention based on disease pathobiology are among the possible applications of ICH therapy. Around the world, traditional plant remedies have been utilized to manage many illnesses. Epidemiology, risk factors, pathogenesis, diagnosis related to machine learning, artificial intelligence, and treatment based on natural remedies and novel approaches for ICH, their modes of action, and the possible advantages.

The pharmaceutical industry stands at the precipice of an AI-driven data revolution, with synthetic patients emerging as a transformative tool to accelerate drug discovery and development while enhancing patient privacy. However, a critical regulatory gap persists: the absence of a standardized basis from leading regulatory bodies for accepting AI-generated patient populations as evidence in regulatory submissions. This manuscript addresses this void by proposing five foundational principles-Representativeness, Utility, Robustness, Privacy Preservation, and Transparency-anchored by the "Fit for Purpose" philosophy. We introduce the operational concept of a "Technical Validation Playbook" to facilitate the first wave of regulatory acceptances for synthetic patient data. We further outline actionable recommendations for regulatory agencies and pharmaceutical sponsors to advance the acceptance of synthetic patient populations through existing qualification and scientific advice mechanisms. By establishing a proactive, principle-based approach, this framework aims to catalyze regulatory-industry alignment and unlock the transformative potential of synthetic patients, particularly for populations with unmet medical needs such as rare diseases where traditional placebo-controlled trials face insurmountable ethical and recruitment challenges.

Large language models (LLMs) have evolved into versatile tools of the 21st century, simplifying repetitive and labor-intensive tasks in everyday life. Here, we aimed to test the feasibility of using different LLMs to assess the greenness of analytical procedures according to the "Analytical GREEnness Metric Approach" (AGREE) by extracting specific data corresponding to the 12 principles of green analytical chemistry from scientific articles. Seven open-access articles on plant natural products using different analytical techniques were evaluated with the five most popular artificial intelligence (AI) tools (ChatGPT, Copilot, Perplexity, Claude, and Gemini), which were tasked to obtain specific data, along with a justification for the selection of this data. Additionally, different versions (basic and advanced) of the same tool (Perplexity and Perplexity Pro), output types (PDF file and link to the online version), and repeatability (three times the same task) were compared. All extracted data were used to calculate AGREE scores, and the final results were compared with those obtained by experts. AI tools were able to assess greenness with a high degree of accuracy, similar to that of trained researchers. Furthermore, a verification/comparison study demonstrated the possibility of critically examining the greenness assessment of developed methods to facilitate standardizing greenness evaluation. The final application of advanced AI tools dedicated to scientific research confirmed the greenness scores, indicating high consistency between all considered evaluation methods.

Aging water infrastructure and the resulting increase in pipe leaks pose significant operational and financial challenges for modern utilities, requiring more accurate tools for failure identification. This study presents a comprehensive benchmarking framework designed to predict pipe failure probability by evaluating a wide array of state-of-the-art classification models, including traditional baselines, tree-based ensembles, and emerging tabular deep learning architectures. The methodology integrates high-resolution datasets with a dedicated evaluation of spatially derived infrastructure indicators to capture the complex environmental and physical drivers of failure. To address inherent class imbalance, the study systematically benchmarks resampling strategies, such as SMOTE, ADASYN, and RUS, to determine whether these techniques truly improve decision-making performance. This assessment is grounded in the application of proper scoring rules, specifically Logarithmic Loss and Brier Score, alongside the introduction of the Area Under the Cost Curve to evaluate the economic implications of predictive performance across varying cost scenarios.

This study presents a systematic comparative evaluation of ten regression-based machine-learning models for day-ahead photovoltaic (PV) energy forecasting under semi-arid climatic conditions. The analysis is conducted using a limited six month dataset (May-October 2024) of real operational production data obtained from a 22 MW grid-connected PV plant in Nakhchivan, Azerbaijan, integrated with key meteorological variables including solar irradiance, air temperature, relative humidity, and wind speed. Linear Regression, Ridge, Lasso, ElasticNet, Support Vector Regression (SVR), Decision Tree, Random Forest, Gradient Boosting, XGBoost, and a Multi-Layer Perceptron (MLP) were benchmarked using mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), coefficient of determination (R2), and five-fold cross-validated R2. The results indicate that regularised linear models-particularly Lasso Regression-provide the most consistent balance between predictive accuracy and generalisation stability under moderate data availability, while Random Forest demonstrates strong cross-validated robustness and the MLP exhibits overfitting behaviour, highlighting sensitivity to limited training data. These findings demonstrate that increased model complexity does not necessarily translate into improved forecasting reliability in data-constrained semi-arid environments. The novelty of this work lies in its unified benchmarking framework that simultaneously evaluates predictive accuracy, interpretability, and generalisation performance using real-world utility-scale operational data. By explicitly linking forecasting reliability with sustainability-oriented planning, the study contributes to more reliable life-cycle cost and emission assessment of large-scale PV systems in emerging renewable-energy markets. The findings offer practical guidance for energy planners and policy-makers seeking transparent, computationally efficient forecasting strategies in semi-arid climates.

Tertiary lymphoid structures (TLS) are organized ectopic immune aggregates within the tumor microenvironment and have been associated with prognosis and response to immunotherapy across multiple cancer types. However, their clinical interpretation cannot be reduced to a simple presence-or-absence assessment, and conventional histopathological evaluation is limited in reproducibility, scalability, and functional resolution. Recent advances in artificial intelligence (AI), digital pathology, spatial omics, and multiplex imaging provide new opportunities to quantify TLS morphology, maturity, spatial distribution, cellular composition, and therapy-associated changes. In this review, we summarize recent AI-based approaches for TLS detection, segmentation, maturity classification, multimodal integration, and non-invasive prediction, while also discussing their validation requirements and current limitations. We further examine TLS maturity as a functional continuum, highlighting the distinction between functionally active mature TLS and morphologically mature but functionally impaired "pseudo-mature" TLS. In addition, we discuss how proximal, distal, and therapy-remodeled TLS may have distinct biological and clinical implications. Finally, we outline a conceptual "quantify-decode-intervene-reassess" framework that links AI-assisted TLS assessment with spatial validation and emerging synthetic-immunology strategies. We emphasize that AI-derived TLS scores and the proposed TLS Functional Vitality Index remain investigational frameworks that require rigorous analytical, biological, and clinical validation before clinical deployment.

Translating nutritional recommendations into practical day-to-day meal choices remains a challenging task, particularly when personalization, nutritional adequacy, dietary diversity, allergies, seasonal availability, and food-group constraints must be simultaneously satisfied. This study presents and evaluates the PLAN'EAT Nutrition Advisor, an Artificial Intelligence (AI)-driven, expert rule-based nutrition recommendation system, designed to generate personalized and nutritionally balanced weekly meal plans aligned with established dietary guidelines and food-group recommendations derived from Sustainable Healthy Diet (SHD) principles. The proposed approach is built upon the PLAN'EAT Expert-Curated Meal Database, a nutritionist-designed repository introduced in this work, comprising 401 expert-curated meals spanning Irish, Spanish, and Hungarian cuisines. Meal-plan generation follows a four-stage pipeline: (1) meal filtering based on country-specific cuisine, seasonality, dietary preferences, and allergies, (2) daily meal plan generation through large-scale sampling and scoring against expert-defined nutritional targets, (3) weekly meal plan assembly and optimization under nutritional and food-group constraints, and (4) diversity optimization to promote dietary variety while preserving nutritional validity. The proposed approach was validated through a large-scale in-silico validation involving 1,000 synthetic user profiles and the generation of 16,000 weekly meal plans corresponding to 112,000 daily meal plans. Adherence to daily and weekly nutritional targets, food-group constraints, and overall meal-plan validity at scale were assessed. In addition, we performed a preliminary descriptive evaluation against state-of-the-art Large Language Model (LLM)-based approaches under two complementary settings: Retrieval-Augmented Generation (RAG) and Supervised Fine-Tuning (SFT). Experimental results demonstrate that the proposed system can efficiently generate personalized, nutritionally compliant, and diverse weekly meal plans while maintaining transparent expert-rule-driven optimization. Furthermore, the descriptive evaluation against LLM approaches suggests that the proposed system achieves more consistent energy and macronutrient adherence under the evaluated conditions.

This study explores the relationship between perceived intrusiveness of AI recommendation and live-stream impulse buying in AI-enabled live-stream shopping. Based on Stimulus-Organism-Response theory and recent studies on AI recommendation source effects, the model considers perceived AI recommendation intrusiveness as algorithmic stimulus, perceived attentional capture as the internal organismic mechanism and live-stream impulse buying as a behavioral response. AI skepticism and trait self-control are further proposed as boundary conditions of the model. A survey was conducted with 417 active Chinese live-stream shoppers, and the data were analyzed using PLS-SEM. The results indicate that perceived intrusiveness of the AI recommendations is directly related to live-stream impulse buying, and indirectly related to live-stream impulse buying through the mediator of perceived attentional capture. AI skepticism attenuates the relationship between perceived AI intrusiveness and perceived attentional capture and trait self-control attenuates the relationship between perceived attentional capture and live-stream impulse buying. The study contributes to the development of the nascent literature on AI recommendations by examining intrusive recommendation experiences in live-stream commerce and by providing a greater understanding of how attention, skepticism, and self-regulatory capacity influence consumers' reactions to the algorithmically personalized shopping cues. Given that the study relies on self-report cross-sectional data, the results are regarded as theoretically-based associations rather than causal effects.