暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
To evaluate comparative outcomes of artificial Intelligence (AI)-based and traditional-based teaching in medical education. The literature search was carried out in CENTRAL, CINAHL, Web of Science, MEDLINE, and EMBASE to identify randomized controlled trials (RCTs) comparing AI-based versus traditional-based teachings in medical education. Estimate of effect size was determined for knowledge score, skills score, and teaching satisfaction score via fixed-effect modelling. Fourteen RCTs enrolling 1116 students who received AI-based teaching (n = 558) or traditional-based teaching (n = 558) were included. The use of AI-based teaching was associated with significantly higher knowledge score (standardized mean difference (SMD): 0.36, 95% CI, 0.24-0.49, P < .00001), skills score (SMD: 0.78, 95% CI, 0.57-0.99, P < .00001), and teaching satisfaction score (SMD: 0.97, 95% CI, 0.66-1.29, P < .00001) compared to the traditional-based teaching. Subgroup analyses with respect to the practical course, theoretical course, duration of course shorter or longer than 1 week were consistent with the main analyses. Meta-regression analysis demonstrated that practical course significantly increased estimate effect for knowledge score (P = .002) and skills score (P = .0001). Meta-analysis of best available evidence (level 1a) indicates that AI-based teaching significantly improves student's knowledge, skills, and satisfactions compared to traditional teaching. However, the available evidence may be subject to publication and reporting bias with high between-study heterogeneity. Future studies should evaluate AI-based teaching in postgraduate settings including speciality and even subspecialties trainings. Key messages What is already known on this topic: Growing evidence from randomized controlled trials demonstrated positive impact of artificial intelligence (AI) in medical education when compared to the traditional approaches. What this study adds: Meta-analysis of best available evidence (level 1a) indicates that AI-based teaching significantly improves student's knowledge, skills, and satisfactions compared to traditional teaching. How this study might affect research, practice, or policy: This study suggests that Future studies should evaluate AI-based teaching in postgraduate settings including speciality and even subspecialties trainings.
Embodied intelligence-artificial intelligence instantiated in physical or virtual bodies that can perceive, communicate, and interact with users and their environments-has been increasingly applied in health care. However, the evidence base remains fragmented because of inconsistent terminology, diverse embodiment forms, and limited synthesis of application domains, target populations, care settings, acceptability, and effectiveness. This fragmentation constrains conceptual clarity and translation into routine health care practice. This scoping review aimed to systematically map the applications of embodied intelligence in health care by classifying embodiment forms, identifying major functional domains, describing target populations and implementation settings, and synthesizing the available evidence on acceptability and effectiveness. This scoping review followed the Arksey and O'Malley framework, with enhancements by Levac et al, and was reported in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) and PRISMA-S (Preferred Reporting Items for Systematic Reviews and Meta-Analyses literature search extension) guidelines. Seven electronic databases were searched from database inception to December 2025, supplemented by gray literature searches and backward citation screening. Eligible studies were primary empirical studies published in English or Chinese that examined embodied intelligence in health care contexts. Two reviewers independently screened records and charted data using a pilot-tested standardized form. Descriptive statistics and thematic synthesis were applied. No formal critical appraisal was conducted because the aim was to map the breadth and characteristics of the evidence base. A total of 83 studies were included. Five embodiment forms were identified: virtual humanoid agents (32/83, 38.6%), physical humanoid robots (32/83, 38.6%), virtual animal-shaped agents (1/83, 1.2%), physical animal robots (13/83, 15.7%), and mechanical robots (5/83, 6%). Applications clustered into 3 functional domains: health management and health education (40/83, 48.2%), mental health promotion (37/83, 44.6%), and physiological health promotion (6/83, 7.2%). Older adults were the most frequently targeted population (45/83, 54.3%). Interventions were mainly implemented in home settings, care homes, laboratories, and hospitals. Twenty-two randomized controlled trials reported generally beneficial effects on health behaviors, mental health outcomes, or cognitive function, although outcome measures were heterogeneous. Twelve studies examined acceptability and generally reported favorable user acceptance. This scoping review provides the first comprehensive synthesis of embodied intelligence in health care using a unified classification of forms, functional domains, populations, and application settings. The findings indicate that embodied intelligence is most mature in "health management and health education" and "mental health promotion," with increasing real-world deployment in home and care home settings. By consolidating fragmented evidence and standardizing terminology, this review offers a practical foundation for clinicians, nurses, and policymakers to support the implementation of embodied intelligence in routine health care. Evidence is limited by heterogeneous outcome measures, many lab-based evaluations, and the absence of formal quality appraisal, underscoring the need for standardized outcome measures, rigorous randomized controlled trials, and longitudinal evaluations to enable scalable and ethically grounded real-world adoption.
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
Tremendous improvement in the use of artificial intelligence has opened new opportunities to analyze the data obtained from electronic health records and imaging. New technologies have tried to overcome obstacles to implement guidelines and recommendations. This review aims to describe the recent progress in the use of machine learning and new technologies in the field of nutrition of the critically ill. Increase in data availability, ability to extract these data and analyze them using machine learning has allowed data scientists together with ICU specialists to improve nutritional screening and assessment and to predict occurrence of obstacles like enteral feeding intolerance or refeeding hypophosphatemia. In addition, new technologies can ensure nasogastric tube positioning and enteral feeding efficacy. Integrated platforms can integrate nutritional needs with most adequate prescriptions and modulate the nutritional administration according to the patient's tolerance and requirements. Analysis of continuous recording of imaging obtained from ultrasound can also predict gastric intolerance. Using machine learning, numerous algorithms and nomograms have been suggested to predict enteral feeding intolerance but validation of these predictions is still required. New technologies integrating energy requirements and delivery of the optimal enteral feeding are very promising.
Biosimilar development is undergoing regulatory change, with growing emphasis on analytical comparability and targeted clinical pharmacology rather than routine comparative efficacy studies. A focused review is therefore needed to understand how global regulatory expectations are evolving, where they are converging, and what this means for future development. This review examines biosimilar regulatory frameworks across agencies, including the FDA, EMA, PMDA, WHO, Health Canada, and ANVISA. It outlines the shift toward a stepwise, totality-of-evidence approach in which analytical similarity, functional characterization, pharmacokinetic/pharmacodynamic comparability, and immunogenicity assessment provide the basis for establishing biosimilarity. It also considers recent regulatory changes supporting a reduced role for comparative efficacy studies, broader use of foreign comparators and reliance pathways, and the emerging application of artificial intelligence in comparability assessment and model-informed development. Relevant regulatory documents and peer-reviewed literature were reviewed to summarize current trends, differences, and likely future directions. For many well-characterized biosimilars, particularly monoclonal antibodies, robust analytical and clinical pharmacology data can address most uncertainty, making routine comparative efficacy studies less necessary in many cases. Greater alignment in comparator policies, study expectations, and responsible use of artificial intelligence could help streamline development, reduce duplication, and improve patient access.
Artificial intelligence is now embedded across the scientific research and publishing ecosystem, influencing discovery, analysis, knowledge translation, authorship, peer review, and editorial workflows. In cardiovascular and biomedical sciences, these developments offer substantial opportunities to accelerate knowledge generation, integrate complex datasets, and improve efficiency and consistency. At the same time, they introduce new risks related to bias, transparency, data integrity, and authorship responsibility, potentially endangering trust in the scientific record. This commentary examines the evolving role of AI in biomedical publishing, with particular attention to generative models and machine learning tools. We review both benefits and limitations, highlight risks such as fabricated content, biased outputs, and erosion of accountability, and discuss why traditional detection approaches are insufficient. Instead, we argue for a shift toward transparency, provenance, and enforceable human responsibility as the core principles guiding AI use, ensuring that AI strengthens rather than undermines scientific rigour and public trust. We outline practical expectations for authors, reviewers, editors, and publishers, with emphasis on reporting standards, reproducibility under rapidly evolving model versions, and the conflict-of-interest implications of AI tooling for the editorial process itself.
We introduce Brain-Artificial Intelligence Interfaces (BAIs) as a new class of Brain-Computer Interfaces (BCIs). Unlike conventional BCIs, which rely on intact cognitive capabilities, BAIs leverage the power of artificial intelligence to replace parts of the neuro-cognitive processing pipeline. BAIs allow users to accomplish complex tasks by providing high-level intentions, while a pre-trained AI agent determines low-level details. This approach enlarges the target audience of BCIs to individuals with cognitive impairments, a population often excluded from the benefits of conventional BCIs. We present the general concept of BAIs and illustrate the potential of this new approach with a Conversational BAI based on electroencephalography (EEG), termed EEGChat. In particular, we show in an experiment with simulated phone conversations that the Conversational BAI enables complex communication without the need to be able to generate language. Our work thus demonstrates the ability of a speech neuroprosthesis to enable fluent communication in realistic scenarios with non-invasive technologies.
The increasing complexity of healthcare systems management requires the development of advanced methodologies to support efficient resource allocation, service delivery, and strategic planning. Artificial intelligence has emerged as an important tool in this domain, offering capabilities to model demand, predict capacity requirements, and inform operational decisions through data-driven insights. This article provides a comprehensive scoping review of AI-based approaches for demand and capacity modelling in healthcare systems. Specifically, it examines AI methods applied to key demand prediction tasks, including outpatient activity, emergency department attendances, hospital admissions and readmissions, and length of stay, as well as capacity planning for beds, workforce, equipment, and other critical resources. The review reports trends in AI models design, learning paradigms, performance reporting, and data usage, and highlights the relationship between demand modelling and downstream capacity predictions. In addition, the paper analyses data infrastructure requirements, commonly used datasets, and the growing role of explainable AI in supporting transparency and trust. Despite recent advances in the field, the integration of AI into healthcare systems faces significant challenges, including concerns related to privacy, ethics, data quality, interpretability, bias, scalability, policy variation, and data interoperability. Addressing these challenges is essential to develop sustainable, fair and resilient healthcare systems. Our review highlights the current state and gaps in the literature, and proposes future directions for advancing the use of AI in healthcare management systems are reviewed.