Generative Artificial Intelligence (GAI) offers promising solutions to long-standing challenges in developing medical imaging methods and applications, including data scarcity, privacy concerns, and class imbalance. However, limited consolidation of publicly accessible synthetic datasets and trained GAI checkpoints restricts reproducibility and benchmarking. This systematic review aims to identify and evaluate such resources and assess their utility in clinical imaging applications. We systematically searched PubMed, IEEE Xplore, and Scopus for studies published between January 2017 and June 2024. Eligible studies generated or used synthetic medical image datasets and publicly released either the dataset or the trained GAI model. Extracted data included imaging modality, dataset characteristics, model architecture, public availability, and evaluation strategy. Of 941 screened records, 35 studies met inclusion criteria, comprising 37 publicly available resources spanning radiology (59%), pathology (16%), ophthalmology (14%), and dermatology (11%). Generative models included generative adversarial networks (73%), diffusion models (21%), autoencoders (3%), and hybrid architectures (8%). As some studies employed multiple model types, these categories are not mutually exclusive. Fifteen (43%) studies provided trained model checkpoints, enabling the generation of task-specific synthetic data. Evaluation methods included quantitative metrics, clinical expert assessment, and downstream performance in classification, segmentation, or detection tasks. Although the reviewed resources support diverse downstream applications, publicly available synthetic datasets and trained models remain scarce. Evaluation strategies vary widely, and the absence of standardized benchmarks limits cross-study comparisons and reliability assessment. To support reproducibility and responsible use of GAI in medical imaging, future work should prioritize the public release of curated synthetic resources, clearer guidance on model selection, and standardized, multi-dimensional evaluation frameworks.
Artificial intelligence (AI) has emerged as a promising tool to address gaps in palliative care access and delivery for adults with serious illness and their family caregivers, particularly in home-based settings where access to specialty care is limited. AI-driven tools, including machine learning, natural language processing, and decision-support systems may enable proactive, personalized, and efficient approaches to addressing several domains of quality palliative care, as defined by the National Consensus Project (NCP), including continuity of care, symptom relief, emotional support, and family caregiver assistance. This scoping review aimed to systematically map the evidence on AI applications designed to assist home-based care for adults with serious illness and their family caregivers, with a focus on their potential role in enhancing palliative care delivery. Six databases were searched from inception to August 2025 using terms related to AI, serious illness, home care, self-care, and caregiving. Eligible studies included peer-reviewed empirical research among adults (≥18 years) with serious illness and/or family caregivers, focusing on AI as a tool to support home-based self-care or caregiver contributions to self-care. Of 1,791 articles screened, 24 met inclusion criteria. Qualitative content analysis identified six themes: (I) personalization and contextual adaptation; (II) multimodal and accessible interfaces; (III) emotional and relational dimensions; (IV) predictive and proactive care; (V) daily routines and care ecosystems; and (VI) equity and access. Personalization emerged as a critical feature, with culturally tailored AI tools improving trust and usability. Limitations of the evidence are that most studies emphasized feasibility, usability, and user experience, over clinical or psychosocial outcomes, limiting insight into AI's real-world impact on palliative care. Evidence was further constrained by heterogeneous designs, language restrictions, and the scarcity of research published in palliative care journals, highlighting the need for more rigorous, context-specific studies. Findings underscore AI's capacity to address core components of palliative care, including predicting and managing symptoms and addressing psychosocial needs. However, the evidence base remains early-stage. Future research should prioritize rigorous evaluation of clinical and psychosocial outcomes, along with co-design with patients, caregivers, and clinicians to ensure alignment between AI innovation and core principles of palliative care.
In 2025, the National Academy of Medicine released an Artificial Intelligence Code of Conduct (AICC). In this commentary, we examine how the AICC introduces governance mechanisms to oversee AI applications and how it can support the ethical development and responsible use of AI in healthcare, paying special attention to the role of nurses. One shortcoming of the AICC is its lack of explicit acknowledgment of nurses, which risks obscuring their indispensable role in the safe, equitable, and effective use of AI in healthcare. We offer practical steps for health leaders to operationalize the AICC. Implementation of the AICC can support robust AI governance in health systems, but nurse expertise must be incorporated. The AICC offers a framework into which nursing perspectives can be embedded to ensure that AI tools positively transform healthcare delivery and enhance the quality and equity of care.
Accurate aortic stenosis (AS) phenotyping requires multimodality imaging which has limited availability. The digital aortic stenosis severity index (DASSi), an artificial intelligence biomarker of AS-related remodeling on single-view 2-dimensional echocardiography, predicts AS progression independent of Doppler measurements. We sought to evaluate the ability of DASSi to define personalized AS progression profiles and to validate its performance as a scalable alternative to multimodality imaging features of functional, structural, and biological AS severity. In the SALTIRE-2 trial (Study Investigating the Effect of Drugs Used to Treat Osteoporosis on the Progression of Calcific Aortic Stenosis 2) of participants with mild or moderate AS, we performed blinded DASSi measurements (probability of severe AS, 0-1) on baseline transthoracic echocardiograms. We evaluated the association between baseline DASSi and (1) disease severity by hemodynamic (peak aortic valve velocity), structural (computed tomography-derived aortic valve calcium score), and biological features ([18F]sodium fluoride uptake on positron emission tomography-computed tomography); (2) longitudinal disease progression (absolute change in peak aortic valve velocity and aortic valve calcium score); and (3) incident aortic valve replacement. We used generalized linear mixed or Cox models adjusted for risk factors and aortic valve area. We analyzed 134 participants (72 [interquartile range, 69-78] years; 27 [20.1%] women) with a mean baseline DASSi of 0.51 (SD, 0.19). DASSi was independently associated with cross-sectional disease severity: each SD increase was associated with higher peak aortic valve velocity (+0.21 [95% CI, 0.12-0.30] m/s), aortic valve calcium score (+284 [95% CI, 101-467] Agatston units), and [18F]sodium fluoride target-to-background ratiomax (+0.17 [95% CI, 0.04-0.31]). Higher DASSi was also associated with disease progression by Doppler (peak aortic valve velocity) and computed tomography (aortic valve calcium score) at 24 months (P interaction for DASSi × time<0.001), and future aortic valve replacement (75 events over 5.5 [interquartile range, 2.4-7.2] years, adjusted hazard ratio, 1.42 [95% CI, 1.11-1.84] per SD). DASSi is associated with functional, structural and biological features of AS severity and predicts disease progression and adverse outcomes. DASSi-enhanced echocardiography may provide an accessible alternative to multimodality AS imaging and serve as a predictive enrichment biomarker in clinical trials. URL: https://www.clinicaltrials.gov; Unique identifier: NCT02132026.
Purpose To evaluate the predictive value of myosteatosis as an opportunistic finding in coronary artery calcium (CAC) CT scans for clinically diagnosed chronic obstructive pulmonary disease (COPD) and compare it with an artificial intelligence (AI)-measured biomarker of emphysema derived from the same scans. Materials and Methods In this prospective study, baseline CAC CT scans and 20-year follow-up data were analyzed. Myosteatosis was defined as the lowest quartile of thoracic skeletal muscle mean attenuation (males < 33.5 HU, females < 27.0 HU). The emphysema-like lung biomarker was quantified as the percentage of lung voxels below -950 HU in CAC CT scans. COPD was identified using the International Classification of Diseases, Ninth Revision, Clinical Modification, and International Classification of Diseases, 10th Revision, Clinical Modification diagnostic codes from hospital discharge records. Hazard ratios (HRs) for COPD were calculated using proportional hazard regression models, comparing the bottom versus top quartiles of myosteatosis and emphysema-like lung measurements. Results Among 5535 participants in the Multi-Ethnic Study of Atherosclerosis (mean age ± SD, 62.2 years ± 10.3, 47.6% males), 396 (7.1%) were diagnosed with COPD over the 20-year follow-up period. Myosteatosis showed a stronger association with COPD than emphysema (unadjusted HRs, 5.98 [95% CI: 4.14, 8.63] and 2.12 [95% CI: 1.61, 2.78], respectively [P < .001]). After adjusting for covariates (age, sex, smoking status, body mass index, race, asthma, physical activity, inflammatory markers, and insulin resistance), the HRs were reduced to 2.74 (95% CI: 1.81, 4.16) and 1.50 (95% CI: 1.12, 2.00), respectively (P = .02). Conclusion AI-measured myosteatosis in CAC CT scans strongly predicted future diagnosed COPD independently of known risk factors. Keywords: Applications-CT, Pulmonary, Thorax, Adipose Tissue (Obesity Studies), Chronic Obstructive Pulmonary Disease, Metabolic Disorders, Myosteatosis, Coronary Artery Calcium Scan, Emphysema, AI-CVD ClinicalTrials.gov: NCT00005487 Supplemental material is available for this article. © The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license.
To systematically map how artificial intelligence (AI) can transform whole-course chronic obstructive pulmonary disease (COPD) management across prevention, diagnosis, treatment and rehabilitation within a 4P (Predictive, Preventive, Personalized, Participatory) medicine framework, and to identify actionable strategies for overcoming current barriers. A systematic search of PubMed, Web of Science and Embase was performed for articles published between January 2021 and June 2025. This review was conducted following the PRISMA guidelines. Forty empirical studies and reviews that applied AI/ML to COPD prevention, early detection, personalised therapy, exacerbation prediction or pulmonary rehabilitation were critically appraised. Data were extracted on technical foundations, data modalities, algorithms, validation metrics and implementation outcomes. AI models integrating multimodal data (imaging, wearables, environmental exposures, genomics) achieved AUC ≥ 0.80 for predicting acute exacerbations up to seven days in advance, were associated with a reduction in emergency visits of up to 98% and a lowering of readmission rates by 25-48%. Screening tools using chest X-ray, CT or smartphone sensors attained ≥90% accuracy for early COPD detection in primary-care settings. Personalised treatment optimisation was linked to a 53% lowering of exacerbation risk in best-responding subgroups. Home-based AI rehabilitation platforms increased adherence by >30% without additional equipment. Key implementation challenges include data heterogeneity, limited explainability, digital divide among older adults and unclear regulatory frameworks. AI is poised to operationalise 4P COPD care, delivering substantial clinical and economic benefits. Future success depends on cross-centre data standards, explainable-AI toolchains, federated learning and inclusive reimbursement policies.
As the healthcare sector increasingly integrates Artificial Intelligence (AI) technologies to improve operational effectiveness, diagnosis, and therapy, the environmental footprint of these innovations has become a growing concern. High energy consumption, electronic waste, and carbon emissions associated with the deployment and training of AI models pose sustainability challenges that must be addressed. This paper investigates the concept and application of Green AI in healthcare, aiming to balance technological advancement with environmental responsibility. This systematic review explores key themes, including green computing practices, the adoption of energy-efficient AI models, and the use of renewable energy sources within healthcare settings. It identifies a range of healthcare applications employing Green AI, highlights emerging trends, and emphasizes the growing importance of environmental awareness in AI development. Furthermore, the study examines enabling tools and techniques, outlines barriers to adoption, and highlights how Green AI can help streamline processes, reduce resource waste, and promote environmentally friendly medical procedures like telemedicine. The review also discusses the significance of policy frameworks, international initiatives, and cross-sector collaboration in promoting environmentally responsible AI deployment. Finally, the paper presents practical implications and outlines future research directions to guide the sustainable evolution of AI in healthcare.
OCCUPATIONAL APPLICATIONSThe adoption of AI in healthcare depends on calibrated trust-trust that matches the system's reliability and context. This review shows that clinicians value workload reduction, explainability, and alignment with clinical judgment, while patients emphasize transparency, fairness, and human-like interaction. Yet, trust is not automatic: performance gains may fail if AI undermines professional autonomy, and explainability reassures novices more than experts in high-stakes tasks. For occupational applications, AI must be designed to reduce cognitive burden, respect user expertise, and adapt to domain-specific needs. Organizations should invest in usability testing, peer and organizational support, and targeted training to foster informed trust. Regulators should enforce transparency and human oversight standards. Ultimately, calibrated trust, avoiding blind reliance or excessive skepticism, is essential to protect healthcare workers and patients while ensuring AI strengthens decision-making and safety. Background: In healthcare, effective and widespread adoption of artificial intelligence (AI) will require users to trust AI. Several studies have explored AI’s accuracy, performance, and implementation challenges, but fewer have examined AI trust from a human factor perspective.Purpose: The objective of this systematic literature review is to explore factors quantitatively affecting trust in healthcare AI from a human factor perspective. Additionally, it investigates the healthcare domains where trust in AI has been studied and the types of AI that have been evaluated.Methods: A systematic review of studies published between April 2014 and April 2024 was conducted to identify factors influencing trust in AI systems within healthcare settings. The search strategy, conducted in PubMed, IEEE Xplore, and Web of Science, utilized keywords related to trust, AI, and healthcare. Studies were included if they reported direct quantitative measures of clinicians’ or patients’ trust in AI systems.Results: A systematic review of 51,085 records identified 29 relevant studies that quantitatively analyzed the factors that influence trust in AI across various healthcare domains. These factors include workload perception, performance expectancy, risk, uncertainty, explainability, transparency, social influence, human representation, and demographic factors. While these factors generally shape trust, their impact varies depending on the application, user experience, and domain.Conclusions: Overall, the review contributes a quantitative, human factors-oriented synthesis of trust in healthcare AI, demonstrates the centrality of trust calibration, and connects these insights to regulatory guidance. These contributions set the foundation for future work to design, evaluate, and govern AI systems that are not just adopted, but appropriately relied upon, across diverse healthcare contexts.
Smart patch healthcare devices are emerging as a distinct user interface in decoding the bidirectional interaction of the five sense organs. Powered by recent advancements in nano-materials, and artificial intelligence predictions, smart patches could understand the immune response of the body by analysing the biofluids, microenvironment and analytes in the five sense organs. These eminent potentials in smart patches, inspired the necessity for a review. Thus, this review aims to bring in to the limelight the current progress in smart patch technologies, highlighting their functions, opportunities and challenges in healthcare applications. A comprehensive review of literature was conducted focusing on smart patches designed for skin, ocular, cochlear, oral, and nasal applications. Further, the review is structured emphasising details on materials used, fabrication methods adapted, sensing mechanisms employed, enabling technologies such as artificial intelligence and Internet of Things. The review analysis revealed that smart patches play a multifaceted role in healthcare applications providing (i) continuous health monitoring, (ii) controlled drug delivery, (iii) supports tissue regeneration and (iv) enables modulation of nerve responses. Further, smart patch integration with Internet of Things (IoT) capabilities enables remote healthcare solutions which benefits both physician and patient communities equally. Despite these progresses, challenges remain in term of biocompatibility of the materials chosen, long-term use and stability of the patch, data security and large-scale manufacturing. Smart patches hold transformative potential in biomedical engineering by bridging biosensing, therapeutic, and digital healthcare domains. This article provides an in-depth review of the current advancements, identifying the existing challenges and emerging opportunities in the field of smart patch research, and thus could guide future research and development. With its broad scope, this review would act as a valuable resource for both researchers and healthcare innovators working towards next-generation biomedical devices.
Chronic obstructive pulmonary disease (COPD) is a leading cause of morbidity and mortality worldwide, with frequent exacerbations of COPD (ECOPD) significantly impacting patient health and health care systems. Predicting ECOPD early would increase patients' quality of life and decrease the economic burden. The advancement of wearable technologies and Internet of Things (IoT) sensors has enabled continuous remote monitoring (RM), offering new opportunities for early ECOPD prediction. However, effectively leveraging wearable data requires robust artificial intelligence (AI) frameworks capable of processing heterogeneous physiological and environmental information. This systematic review aims to provide a comprehensive overview of both hardware and software solutions for predicting ECOPD using RM. From the reviewed literature, we first focus on key physiological and environmental variables essential for COPD monitoring that can be extracted from wearables and IoT sensors. Second, we describe the wearable and IoT devices currently deployed in COPD management. Finally, we review machine learning, including deep learning models, used for ECOPD prediction, discussing limitations for real-world implementation. By bridging AI-driven data processing with real-world sensor applications, this review aims to outline the current landscape, existing challenges, and future directions for developing effective RM solutions for ECOPD predictions. A comprehensive search was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to identify studies using AI or machine learning techniques for predicting ECOPD in in-home contexts. This review identified 26 studies that met the inclusion criteria. Twenty studies aimed at predicting or detecting exacerbations at the onset. The variables tracked most frequently were heart rate (n=9), peripheral oxygen saturation (n=9), and symptoms (n=8). Daily or weekly sampling was most common (n=14). Most studies (n=13) applied machine learning models-primarily random forest (n=5), CatBoost (n=2), decision trees (n=2), and support vector machines (n=2). Deep learning was used in 3 papers, while the remaining applied rule-based logics and probabilistic models. Wearables and IoT were used in only 6 out of 20 studies. Six papers analyzed changes in vital parameters during prodromal phases, defined as the period shortly before the onset of an exacerbation. Three studies collected data continuously, 2 daily, and 1 compared once-daily versus overnight monitoring; 4 of these 6 used wearable devices. Overall, current evidence highlights the potential of continuous monitoring of physiological and environmental variables for early ECOPD prediction, offering advantages over questionnaires or once-daily measurements. While wearables and IoT devices show promise, their use remains limited. Many studies rely on balanced datasets that do not mirror real-world exacerbation patterns and lack external validation across diverse populations. Future research should emphasize large-scale validation, integration of multimodal data, and translation of AI models into clinically feasible tools to enable timely intervention and improve COPD management.
In the internet of medical things, data primarily exhibits time-series and streaming characteristics, featuring typical attributes such as large-scale volume, high transmission rates, and significant heterogeneity. Given these data properties and the application requirements of medical scenarios, the development of specialized data platforms tailored to these needs holds considerable research significance and practical value. This study innovatively proposes the internet of medical things data platform solution based on a cloud-edge-end architecture, and elaborates on its architecture, functions, and implementation effects. The edge side is responsible for streaming data access, storage, and computation; the cloud side encompasses three layers of services: resources, data, and applications, constructing a data lake to provide data analysis services. This study has been implemented in PLA General Hospital for verification. From 2021 to 2024, 263 medical devices have been connected accumulatively, with a total data volume of 24.07 TB and stable operation within 4 years. In the performance stress test, the platform achieved the data access throughput of 23.91 MB/s and the data storage efficiency of 30.98 MB/s. These results demonstrate the feasibility of the architecture platform. This study has engineered and successfully applied the cloud-edge-end architecture in complex internet of medical things scenarios, addressing challenges such as heterogeneous protocol compatibility of medical devices, real-time response to clinical operations, and large-scale storage and application of the internet of things data. The established data platform provides a solid data foundation for smart medical applications and holds significant value for the research of medical artificial intelligence and the construction of future smart hospitals. 医疗物联网中数据主要为时序与流式形态，具有规模大、传输速率高以及异构性强等典型特征。鉴于这些数据特性与医疗场景的应用需求，开发与之适配的专用数据平台，具有重要研究意义和应用价值。本研究创新性提出了基于云—边—端架构的医疗物联网数据平台解决方案，并对其架构、功能与实施效果进行了阐述。其边缘侧负责流数据接入、存储与计算；云侧涵盖资源、数据与应用三层服务，构建数据湖，提供数据分析服务。本研究在中国人民解放军总医院进行了实施效果验证，自2021—2024年累计接入263台医疗设备，数据总量24.07 TB，系统持续稳定运行四年；性能压力测试中平台数据接入吞吐量23.91 MB/s，数据存储效率30.98 MB/s，其结果证明了该架构平台的可行性。本研究将云—边—端架构在复杂医疗物联网场景下进行了工程化落地和成功应用，解决了医疗设备异构协议兼容、临床业务实时响应以及物联网数据大规模存储应用等难题，建立的数据平台为智慧医疗应用提供了坚实的数据底座，对医学人工智能的研究和未来智慧医院的建设具有重要价值。.
The pace of surgical innovation appears ever faster. Innovation is being freed from the design constraints of the opposable digits of a surgeon's hand through the use of programmable binary digits. Surgeons must be the drivers of change and central to the application of innovations. We should collaborate with industry, engineers and scientists to think out of the box but must consider also expense, environmental impact, equity, and ethics. But we should not be blinded by shiny technology: innovation without impact is mere noise. The ultimate considerations are the diagnosis and management of surgical disease, of improving the care of our patients. Expert surgeons, scientists and engineers across the world were identified and invited to describe areas of innovation within surgery. They were given free rein to review their areas of expertise and to discuss both current and future applications of technology within surgical care. The Commission spans multiple surgical specialties and scientific domains. It reviews translational genomics, including the role of ctDNA, alongside microbiomic and proteomic applications in improving the diagnosis, treatment and monitoring of surgical disease. Applications to enhance surgical procedures are described, from medical micro/nanorobots for minimally invasive interventions, sensory-enriched surgery with visual optimization and molecular image-guidance to intelligent and semiautomated instruments. The expansion and broad influence of artificial intelligence in surgical writing, training and simulation, diagnosis and robotics is widely described. The role of surgical innovation and technology in driving personalized care for benign and malignant surgical disease from genomic profiling to bespoke surgical and non-surgical treatment pathways and surveillance is considered. The future of surgery is poised to become more precise, personalized, and effective. Collaboration with engineers, data scientists, and industry partners not only represents an exciting opportunity for surgeons to participate in team science but is critical to focus innovation goals on optimizing patient care and outcomes.
Breast cancer is a leading cause of mortality and morbidity among females worldwide. As part of the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2023, we provided an updated comprehensive assessment of the epidemiological trends, disease burden, and risk factors associated with breast cancer globally, regionally, and nationally from 1990 to 2023. Breast cancer incidence, mortality, prevalence, years lived with disability (YLDs), years of life lost (YLLs), and disability-adjusted life-years (DALYs) were estimated by age and sex for 204 countries and territories from 1990 to 2023. Mortality estimates were generated using GBD Cause of Death Ensemble models, leveraging data from population-based cancer registration systems, vital registration systems, and verbal autopsies. Mortality-to-incidence ratios were calculated to derive both mortality and incidence estimates. Prevalence was calculated by combining incidence and modelled survival estimates. YLLs were established by multiplying age-specific deaths with the GBD standard life expectancy at the age of death. YLDs were estimated by applying disability weights to prevalence estimates. The sum of YLLs and YLDs equalled the number of DALYs. Breast cancer burden attributable to seven risk factors was examined through the comparative risk assessment framework. The GBD forecasting framework was used to forecast breast cancer incidence and mortality from 2024 to 2050. Age-standardised rates were calculated for each metric using the GBD 2023 world standard population. In 2023, there were an estimated 2·30 million (95% uncertainty interval [UI] 2·01 to 2·61) breast cancer incident cases, 764 000 deaths (672 000 to 854 000), and 24·1 million (21·3 to 27·5) DALYs among females globally. In the World Bank low-income group, where a low age-standardised incidence rate (ASIR) was estimated (44·2 per 100 000 person-years [31·2 to 58·4]), the age-standardised mortality rate (ASMR) was the highest (24·1 per 100 000 [16·8 to 31·9]). The highest ASIR was in the high-income group (75·7 per 100 000 [67·1 to 84·0]), and the lowest ASMR was in the upper-middle-income group (11·2 per 100 000 [10·2 to 12·3]). Between 1990 and 2023, the ASIR in the low-income group increased by 147·2% (38·1 to 271·7), compared with a 1·2% (-11·5 to 17·2) change in the high-income group. The ASMR decreased in the high-income group, changing by -29·9% (-33·6 to -25·9), but increased by 99·3% (12·5 to 202·9) in the low-income group. The increase in age-standardised DALY rates followed that of ASMRs. Risk factors such as dietary risks, tobacco use, and high fasting plasma glucose contributed to 28·3% (16·6 to 38·9) of breast cancer DALYs in 2023. The risk factors with a decrease in attributable DALYs between 1990 and 2023 were high alcohol use and tobacco. By 2050, the global incident cases of breast cancer among females were forecast to reach 3·56 million (2·29 to 4·83), with 1·37 million (0·841 to 2·02) deaths. The stable incidence and declining mortality rates of female breast cancer in high-income nations reflect success in screening, diagnosis, and treatment. In contrast, the concurrent rise in incidence and mortality in other regions signals health system deficits. Without effective interventions, many countries will fall short of the WHO Global Breast Cancer Initiative's ambitious target of achieving an annual reduction of 2·5% in age-standardised mortality rates by 2040. The mounting breast cancer burden, disproportionately affecting some of the world's most vulnerable populations, will further exacerbate health inequalities across the globe without decisive immediate action. Gates Foundation, St Jude Children's Research Hospital.
The advancement of artificial intelligence (AI) technologies has resulted in the proliferation of novel applications in various fields, including nutrition. One of the most notable applications involves AI-generated and guided diet plans. The present study evaluates diet plans generated by various AI tools (e.g., ChatGPT, Gemini, DeepSeek, etc.) for individuals with different health profiles using a multi-criteria decision-making (MCDM) framework. A series of authentic client scenarios were formulated on the basis of anonymized clinical cases that had been provided by a registered dietitian. These scenarios incorporated medical history, lifestyle habits, dietary patterns, and other relevant factors. For each client's profile, a set of standardized prompts were submitted to different AI tools to generate comparable diet plans. The resulting diet plans were evaluated based on several main and sub-criteria, including appropriateness, feasibility, nutritional adequacy, degree of personalization, ethical compliance, reproducibility, and linguistic clarity. The evaluation employed MCDM methods, namely LBWA for weighting, COPRAS, and PROMETHEE-I/II for ranking. The findings indicate that GPTPLUS demonstrated the highest overall ranking; DeepSeek exhibited consistent second-tier performance; and mid-tier models (GPT-4.0, GPT-4.5, Grook3) exchanged positions depending on the scenario and method. The results at the criterion level were found to be aligned with clinical priorities. Moreover, Claude's refusal to formulate a dietary plan for a client under the age of 18 indicates a paucity of standardized ethical guidelines governing the utilization of AI tools. The findings emphasize the potential of AI as a supportive tool in healthcare services, while concurrently addressing ethical considerations and practical limitations.
Ontologies support transparent and reproducible conceptual modeling in Health Technology Assessment (HTA), but their population remains resource-intensive and reliant on expert input. This study evaluates the feasibility, reliability, and methodological implications of using generative artificial intelligence (GenAI) to populate ontology individuals for HTA applications. A factorial experimental framework was developed using the Ontology for Simulation Modeling (OSDi) and three HTA-relevant use cases of varying complexity. Two GenAI systems were evaluated under multiple experimental conditions, including prompting strategy, serialization format, and provision of supporting information. Generated ontology individuals were validated by an HTA expert and assessed across four quality dimensions: consistency, relevance, completeness, and adequacy. Multivariate and regression analyses were conducted to examine the effects of experimental factors on quality outcomes and hallucination likelihood. GenAI systems successfully generated ontology individuals across use cases, although performance varied by quality dimension and experimental condition. Iterative prompting significantly improved completeness, while serialization format strongly influenced reliability, with Turtle serialization associated with substantially lower hallucination likelihood compared with XML. Other factors showed dimension-specific effects, highlighting the multidimensional nature of ontology quality. Errors occurred more frequently in structurally complex ontology components, suggesting a relationship between ontological complexity and generative performance. GenAI-assisted ontology population can enhance the efficiency and scalability of HTA conceptual modeling, enhancing the agility of HTA agencies in exploratory phases. Its effective use requires structured prompting, appropriate representation formats, and expert validation. Further research should evaluate its impact on HTA decision modeling workflows and validation frameworks.
Paper-based microfluidic analytical devices (μPADs) form a hydrophilic channel by patterning a hydrophobic barrier, and use capillary action to achieve pumpless fluid control. As point-of-care testing (POCT) platforms, they have significant advantages such as low cost, ease of operation, and low sample consumption compared with traditional microfluidic chips. They are especially suitable for biomarker detection in resource-limited areas and home healthcare settings. To overcome the high cost and environmental instability of traditional natural enzymes as biorecognition elements, redox nanozymes with peroxidase (POD)-like activity have become a stable and economical alternative. Through the synergistic combination of physical adsorption and in situ growth techniques, nanozyme-μPAD constructs offer expedited, user-friendly detection capabilities for a broad spectrum of analytes, encompassing disease biomarkers, pathogenic agents, and environmental pollutants. This article reviews the research progress of the functional transformation of μPADs through the integration of nanozymes, systematically discusses the categories, design, screening, and regulation strategies of nanozymes, and elaborates on their catalytic mechanism in signal amplification. The article focuses on the key applications of nanozyme-functionalized μPADs in the fields of biomedical diagnosis and food safety detection, and outlines future development trends, existing challenges, and potential research directions of this technology.
Artificial intelligence (AI) is increasingly run on high-density computing infrastructure, yet its environmental footprint is still assessed mainly through electricity use and associated greenhouse-gas emissions. A critical, less visible dimension is water: AI infrastructure consumes freshwater through evaporative cooling, indirect water use in electricity generation, and water-intensive semiconductor manufacturing. Projections suggest AI's global water footprint could reach 4.2-6.6 billion cubic meters annually by 2027. Many data centers are located in water-stressed regions. While technologies, including cold-climate siting, natural water body cooling, waterless designs, and waste heat recovery, can reduce on-site demand, their deployment remains limited. This work introduces "digital water sobriety" as a governance framework linking evaluation of which AI applications justify freshwater consumption, water conscious siting, and mandatory facility-level water use transparency. Achieving water-sustainable AI demands not merely technological optimization but fundamental policy reform integrating water constraints into computational infrastructure planning.
Early diagnosis of hepatocellular carcinoma (HCC) depends on rapid, sensitive, and reliable immunoassays; however, conventional ELISA techniques are often limited by complex manual operations, prolonged assay durations, and insufficient sensitivity, which significantly impede the detection of low-abundance biomarkers at the early disease stage. In contrast, microfluidic technology minimizes reagent consumption, enables precise spatiotemporal control of reactions, and provides a versatile platform for developing magnetically guided, multi-step immunoassay systems suited for portable, on-site screening applications. Herein, we have developed a portable microfluidic device with magnetic control, wherein an algorithm-driven stepper motor moves the microfluidic chip across a fixed magnetic-field array, thereby creating tunable local magnetic fields. This design integrates digital algorithmic control, mechanical actuation, the microfluidic chip, and functionalized magnetic beads into a seamless control chain, enabling full automation of bead capture, bead transport, and a sequence of immunoreaction steps without manual intervention. Using alpha-fetoprotein (AFP) as a model biomarker, the device achieved a detection limit of 119.43 ng/mL in the colorimetric measurement mode, and an impressive 3.66 ng/mL in the chemiluminescent mode. The results demonstrate the feasibility of automated, multi-step immunoassay processing in a compact point-of-care format, with potential applications in rapid and sensitive biomarker screening for early disease diagnosis. Importantly, this platform substantially fulfils pivotal deficiencies in HCC screening by markedly abbreviating assay duration and attenuating procedural complexity, thereby enhancing accessibility within resource-limited milieus. By facilitating antecedent detection at a juncture when curative interventions are optimally efficacious, it holds considerable promise for elevating patient prognoses, prolonging survival trajectories, and democratizing advanced diagnostic capabilities across broader populations.
This study presents a comprehensive review of the emerging role of Generative Artificial Intelligence (GenAI) in environmental assessment and sustainability analysis. Positioned within a new paradigm of environmental management, GenAI redefines traditional static models through dynamic, generative, and participatory approaches that integrate data synthesis, scenario modeling, and governance insight. Using a Systematic Literature Review (SLR) guided by the CIMO (Context-Intervention-Mechanism-Outcome) framework, this paper identifies and analyzes 182 scholarly and technical publications published between 2015 and 2025. The review synthesizes developments across key GenAI architectures-Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Transformer-based Large Language Models (LLMs), and Diffusion Models-and evaluates their applications in synthetic data generation, scenario simulation, remote sensing, predictive analytics, and public engagement. The findings reveal that GenAI holds significant potential to address data scarcity, enhance model scalability, and promote participatory and interdisciplinary decision-making, while also presenting challenges related to interpretability, data bias, validation, environmental footprint, and ethical governance. To guide responsible implementation, the study proposes a three-tier framework emphasizing technical fidelity, transparency, and governance alignment. The implications underscore that effective integration of GenAI into environmental management requires hybrid modeling, participatory data governance, and sustainable AI infrastructures to ensure transparency, accountability, and equity. Collectively, this work advances an evidence-based understanding of how GenAI can underpin a data-driven, inclusive, and ethically responsible paradigm in environmental assessment.
Sensor technology has emerged as a transformative tool for point-of-need and portable quality control and safety assessment of Traditional Chinese Medicine (TCM) products. Although separation and detection technologies have improved, there is still a strong need to critically evaluate the entire analytical workflow from sample preparation to data interpretation to meet regulatory and routine monitoring requirements. This review critically evaluates the complete analytical workflow, from sample pretreatment to intelligent data interpretation, necessary to bridge the gap between laboratory prototypes and real-world deployment. Additionally, it provides a systematic summary of the mechanisms and applications of common types of sensors, utilising nanomaterials and molecular recognition elements to overcome the specific challenge of complex TCM matrices. A particular focus is placed on the emerging roles of Artificial Intelligence (AI) and Machine Learning (ML) in resolving high-dimensional spectral data and enhancing pattern recognition accuracy. Furthermore, we critically analyze current bottlenecks, including matrix interference, the lack of standardized validation protocols, and the reliability of portable devices in uncontrolled environments. Future directions are highlighted towards the development of self-validating, cloud-integrated (IoT), and eco-friendly sensing systems, aiming to support the global standardization and safety assurance of TCM products.