搜索结果：AI & society

Generative artificial intelligence (AI) has emerged as a transformative tool for creating high-quality visual materials in medical research and education. In pediatric neurosurgery, where ethical and legal constraints limit the use of real patient photographs, AI-assisted illustrations offer significant potential. However, concerns regarding clinical accuracy, intellectual property, and the protection of vulnerable pediatric patients necessitate rigorous oversight. We present a human-in-the-loop workflow that integrates generative AI with vector-based digital editing to produce scientifically accurate and ethically grounded medical illustrations. We reviewed current AI usage policies from major medical journals, including the International Committee of Medical Journal Editors (ICMJE) and the Journal of Korean Neurosurgical Society (JKNS). To demonstrate practical application, we developed illustrative examples for conditions such as sacral dimple, Crouzon syndrome, and Down syndrome using clinician-led sketches and AI-assisted refinement. Vector-based workflows facilitate the transformation of AI-generated raster drafts into editable, high-resolution graphics, allowing clinicians to correct "hallucinations" and ensure anatomical precision. While most journals prohibit listing AI as an author, they permit its use for conceptual figures provided there is transparent disclosure of the tools and prompts used. Our proposed workflow emphasizes that AI should function as a "constrained assistant" rather than an autonomous creator, ensuring that the final output remains non-identifiable and respectful of pediatric patients' dignity. Generative AI tools can significantly enhance visualization in pediatric neurosurgery when governed by strict ethical and technical safeguards. Adherence to journal policies and the maintenance of human-directed validation are essential to uphold scientific integrity and patient privacy in the era of AI-assisted publishing.

Social media platforms facilitate global discourse on the application of artificial intelligence (AI) in healthcare. Nevertheless, there is a paucity of longitudinal analyses of digitally mediated discussions. To investigate the evolution of global English-language-dominated discourse on #AIinHealthcare over a three-year period on X (formerly Twitter). Using Fedica analytics, we analysed 57,880 tweets by 17,991 distinct users across 141 countries from 1 November 2022 to 1 November 2025. This analysis focused on English-language-dominant discourse around #AIinHealthcare (96.9% English), acknowledging hashtag-specific selection bias and linguistic limitations. This study used publicly available anonymised data and followed the ethical guidelines for social media research. The #AIinHealthcare garnered 39.2 million impressions, with significant contributions from high-income countries, notably the United States (40.7%) and Canada (21.0%), as well as India (13.4%; a rapidly expanding economy), collectively accounting for 75.1% of tweets and reflecting hashtag-specific, geographically concentrated engagement. This peaked in mid-2023 and stabilized lower by mid-2025. English was the predominant language of the discourse (96.9%). The community consisted of 74.9% grassroots users with fewer than 1,000 followers, suggesting genuine participation beyond elite influencers. Total engagement reached 72,625 interactions, primarily passive, comprising 68.1% likes, 19.4% retweets, 10.3% replies, and 2.1% quote tweets. Hashtag co-occurrence patterns, supported by qualitative inspection of exemplar tweets, indicated majorly five distinct clusters: foundational technical topics (#GenerativeAI, #ChatGPT, #LLMs) peaked after November 2022; clinical application themes emerged across disease-specific specialties (#Oncology, #Cardiology, #MentalHealth); healthcare implementation themes addressed practical integration (#DigitalHealth, #Telemedicine, #EHR); governance and ethics themes gained prominence (#ResponsibleAI, #AIEthics, #ExplainableAI, #DataPrivacy); and professional integration themes fostered learning communities (#MedTwitter, #MedicalEducation). Sentiment was predominantly neutral (95%), with positive (3%) and negative (2%). Monthly tweets peaked in mid-2023 at 1,600-1,800 before declining to 750-900 per month by June 2025. High-engagement content linked AI to practical applications, governmental initiatives, and clinical breakthroughs. English-language-dominated discourse around #AIinHealthcare reveals hashtag-specific maturation from technical enthusiasm to governance and implementation focus. However, platform access restrictions in countries such as China and Russia may skew geographic representation. Disparities in sustainability discourse remain prevalent.

In recent years, artificial intelligence (AI) has made significant strides, gaining traction across various domains, including clinical medicine. The integration of AI as a decision-support tool introduces complexities, particularly in ensuring patient safety and upholding clinical accountability. This is especially pertinent in clinical neurophysiology (CNP), where AI shows promise in enhancing the interpretation of neurophysiologic data from modalities such as electroencephalography, electromyography, and others. Recognizing the potential and inherent challenges of AI integration, the International Federation of Clinical Neurophysiology (IFCN) has set forth guidelines to steer the responsible development, evaluation, and application of AI technologies in CNP. The IFCN's core position on AI in CNP emphasizes improving healthcare outcomes, prioritizing patient-centered care, maintaining transparency in AI-generated interpretations, and supporting, rather than replacing, clinical expertise. As AI technology evolves, the IFCN stresses that models implemented in clinical practice, especially when lacking supervision by experts must meet stringent standards for accuracy, safety, and quality control, and that AI cannot substitute expert care without adequate oversight. The successful integration of AI in CNP hinges on dataset diversity, transparent and ethical training practices, and balancing model complexity with real-world validation. Continuous performance monitoring and clinician feedback are vital to maintaining AI system reliability over time. The IFCN envisions a future where AI ethically enhances CNP practice through multidisciplinary collaboration, focusing on patient safety, clinical integrity, and scientific rigor. Properly implemented, AI can revolutionize CNP by offering real-time decision support, personalized interventions, and remote monitoring, thus advancing the quality of care in CNP.

This paper explores the connection between caring for the land and caring for ourselves, specifically as a response to, and within the context of, adverse mental health among Hawaiian farmers. After providing a brief, empirically informed overview of the farmer mental health crisis, we turn our attention to the native Hawaiians' rich and sophisticated tradition of environmental philosophy, which is underpinned by the land-care ethic of mālama 'āina. We show that the Hawaiian environmental ontology is relational, in that nature and people are intimately and metaphysically interconnected, and we argue that because of this metaphysical embeddedness of people within the land, in the Hawaiian worldview there is no clear distinction between land-care, community-care, and self-care. Moreover, we posit that traditional Hawaiian thought frames agricultural economics, ecology, and wellbeing as complementary concepts, rather than mutually exclusive modalities. We aim to illustrate that this holistic approach to land and community care means that Hawai'i is perhaps uniquely positioned to address the present farmer mental health crisis and, consequently, we conclude that the reintegration of native Hawaiian philosophy into contemporary Hawaiian society would be likely to precipitate improvements in the mental health outcomes of (native and non-native) Hawaiian agricultural workers, and Hawaiian society at-large.

Heart failure (HF) remains one of the most prevalent and burdensome chronic conditions worldwide, with rising incidence and poor prognosis despite major therapeutic advances. Effective management requires a comprehensive, multidisciplinary approach that ensures diagnostic accuracy, therapeutic optimization, and continuity of care from hospital to community settings. This consensus document outlines an integrated model for HF care in Italy, developed to improve patient outcomes and resource efficiency. HF should be diagnosed and phenotyped through clinical evaluation, electrocardiography, echocardiography, biomarkers, and, when indicated, advanced imaging. Guideline-directed medical therapy, including sodium-glucose cotransporter-2 inhibitors, angiotensin receptor-neprilysin inhibitors, β-blockers, and mineralocorticoid receptor antagonists, remains the cornerstone of treatment across the ejection-fraction spectrum. Device therapy, cardiac rehabilitation, and iron supplementation complement pharmacological management. Telemedicine and remote monitoring enable early detection of clinical deterioration and strengthen hospital-community integration, while multidisciplinary HF clinics coordinate pharmacological care, rehabilitation, and patient education. Structured training of healthcare professionals and caregivers, along with therapeutic education programs, enhances adherence, empowers patients, and promotes self-management. Community-based associations further support cardiovascular prevention through educational and screening initiatives. The integration of hospital, community, and digital health resources, combined with continuous professional training and patient empowerment, represents a sustainable and effective model to improve survival, reduce hospitalizations, and enhance quality of life for patients with HF.

While new expensive medicines often offer substantial benefits to patients, they can carry inherent drawbacks such as uncertainty regarding efficacy translating into effectiveness, safety and rational use, as well as a substantial financial burden on society and/or patients. Rational use of medicines aims to maximize effectiveness whilst minimizing side effects, patient burden, and societal costs. In the Netherlands, initiatives aiming to improve the rational use of expensive medicines are being carried out with increasing frequency and in a programmatic manner. This review identified the strategies used, and includes a structured approach for their application during the medicine's use in clinical practice. Rational use initiatives, driven by clinicians and pharmacists, and funded by the Dutch Ministry of Health and the Dutch health insurers were evaluated for strategies that aim to improve the rational use of expensive medicines. In addition, a non-systematic narrative review was carried out through searches in Google, Google Scholar, Pubmed, the Artificial Intelligence (AI)- tools Global Campus and Evidence Hunt to identify additional strategies. Identified strategies were categorized by assessing whether they aimed to address the efficacy-effectiveness gap or to reduce the side effects and societal burden of expensive medicines. Thirteen strategies to improve rational use were identified. Two strategies were identified that aim to address the efficacy-effectiveness gap: optimize patient selection and generating evidence on clinical endpoints. 11 additional strategies that aim to reduce side effects or societal burden were identified: dose reduction, personalized dose optimization, interval lengthening, shortening of the treatment duration, biosimilar/generic drug use, non-medical drug switching, reduction of additional non-medication costs, reduction of drug wastage, switching the route of administration, boosting (improve drug exposure and/or reduce the dose by influencing pharmacokinetic parameters through co-interventions), and optimization of medication adherence. Rational use of expensive medications is essential as part of a drug's life cycle and can benefit patients as well as society. The framework and strategies described in this overview provide guidance for the future rational use of expensive medicines, both for those already in use and for those newly introduced.

Advances in computer vision have fueled the development of artificial intelligence (AI)-based algorithms for pathology. AI-assisted approaches may streamline the diagnostic workflow and reduce variability. To assess the impact of an AI-assist model for human epidermal growth factor receptor 2 (HER2) scoring on pathologist reproducibility and accuracy and to understand pathologist-model interactions. An AI-Assist algorithm for HER2 scoring, AI-Measurement of HER2 (AIM-HER2), was developed to generate slide-level scores of HER2 immunohistochemistry (IHC) aligned with guidelines from the American Society of Clinical Oncology/College of American Pathologists. AIM-HER2 was assessed in a retrospective reader study wherein HER2-trained pathologists (n = 20) scored breast cancer cases (n = 200) with and without model assistance using a 2-cohort crossover design with a 3-week washout. A separate panel of expert pathologists (n = 5) provided manual reference scores. As an AI-assist tool, AIM-HER2 improved interrater agreement both overall and at the 0/1+ and 1+/2+ cutoffs and significantly increased positive percentage agreement at the 0/1+ and 1+/2+ cutoffs. Pathologists displayed a wide range of model override rates, and the quality of these overrides was correlated with each pathologist's manual accuracy. Measurements of AIM-HER2 accuracy were highly dependent on reference panel composition. The use of AI-assist tools, such as AIM-HER2, for scoring HER2 IHC in breast cancer may improve pathologist reproducibility and accuracy, particularly at the 0/1+ and 1+/2+ cutoffs. However, improved consistency of pathologist interpretation of AI-assisted IHC scoring guidance may be necessary for AI-assist tools to reach their full potential.

Responding to the lack of academic research on how young people are impacted by deepfake sexual abuse or how schools should address these issues, this paper explores levels of awareness of AI technology and sexualized deepfakes in UK schools and how schools are responding to these newly emergent harms. Drawing on interviews with students and teachers from eight schools across the UK, we found that teachers and students express uncertainty about how AI deepfake technology works. Some teachers underestimated how easy the technology is to use, and they lacked uniform comprehension that sexualized deepfakes should be treated the same way as non-consensual nudes, leading to inconsistency and variations in school responses. Students similarly lacked basic literacy about AI, equating AI with LLMs like ChatGPT, and even though sexualized deepfakes were occurring in their school contexts, students reported having received no explicit education on the topic. Educators and students connected sexualized deepfakes to a rise in misogyny via social media influencers, with some of the students and teachers calling for more education on AI, sexual violence, and consent at earlier ages. We advance the concept of AI-generated image-based sexual abuse, arguing that these harms should be understood as elements of technology-facilitated gender-based violence (TFGBV). We argue this framing is necessary to support systematic understandings of this issue and develop appropriate school responses. Our discussion offers recommendations for improving AI literacy, including preventative AI education that engages critically with AI harms and supports victims.

Responsible adoption of artificial intelligence (AI) in cardiology remains uneven. We aimed to map knowledge, attitudes, beliefs and practices among cardiovascular professionals in France and to identify levers for implementation. We conducted a national multiprofessional survey across cardiovascular care from 4 December 2024 to 1 March 2025. Prespecified outcomes included regular use in practice, confidence in diagnostic outputs, performance expectations, training needs, and social influence. Seven hundred fifty-six professionals completed the survey (58.2% cardiologists, 24.3% allied-health professionals, 17.8% other professionals; median age 37 years; 46.7% women). AI use was reported as regular (≥ weekly) by 23%, occasionally by 40%, and none by 37%; only 7.8% had formal AI training. Use concentrated on AI-assisted imaging (32%) and patient monitoring/management (18%). The most valued benefit was improved diagnostic accuracy (29%); leading concerns were algorithmic bias (29.9%) and data privacy (28.2%). Explainability increased confidence (among cardiologists, high confidence 64% in therapeutic contexts vs. 84% with explanations). In multivariable analyses, prior training (aOR 3.22, 95% CI 1.60-6.55), research involvement (2.94, 1.90-4.58), and male sex (1.64, 1.05-2.59) were associated with higher use, while age > 40 years was associated with lower use (0.62, 0.40-0.96). Allied-health professionals reported lower social influence and training needs. Adoption of AI in cardiology remains limited, and four levers emerged for responsible scale-up: Training (education), Explainability (transparent outputs), Integration (workflow embedding), and Accompaniment (peer support, evaluation). These priorities should guide education, governance, and procurement strategies.

The increasing adoption of virtual reality (VR) in medical education offers substantial opportunities for immersive, practice-oriented training that complements traditional teaching methods. In particular, VR enables repeated, risk-free exposure to complex clinical scenarios and supports the development of clinical reasoning, communication skills, and procedural competence. However, implementing VR-based courses remains challenging due to high development costs, technical complexity, and the need for close interdisciplinary collaboration. This tutorial presents key insights and best practices from the medical tr.AI.ning project, a 3-year interdisciplinary initiative funded by the German Federal Ministry of Education and Research. The project's objective was to develop an artificial intelligence (AI)-supported, VR-based training platform that allows medical students to practice clinical decision-making in immersive, interactive scenarios. The paper is structured as a tutorial and offers recommendations for planning, developing, and integrating VR courses into medical curricula. Each recommendation is illustrated with concrete examples from our project, serving as a practical blueprint to guide educators and developers in applying these guidelines in their own contexts. Successful implementation of a VR project in medical education requires strategic planning and collaboration, starting with a thorough identification of curricular gaps that VR can address and a clear justification of its added educational value. An interdisciplinary consortium that combines expertise from medical didactics experts, computer science, and design is essential to ensure the development of high-quality, pedagogically sound simulations and intuitive user interfaces. Key factors for success include defining specific learning objectives aligned with competency-based frameworks; iterative development with continuous feedback from medical experts, educators, and students; and structured pilot testing with systematic collection of quantitative and qualitative data to assess usability, immersion, and learning outcomes. Early engagement and walkthroughs with end users help identify practical challenges and inform iterative improvements. A dedicated authoring tool within the project allows medical teachers to create and adapt VR scenarios without prior technical experience, supporting the scalability and sustainability of the approach. Effective project management frameworks facilitate collaboration, clear task allocation, and adaptive progress throughout development. Additionally, considerations for hardware selection, technical infrastructure, and sustainable dissemination strategies, including open-access publications, project websites, and professional networking, are crucial to ensure long-term viability and broad adoption across institutions. By combining a tutorial format with practical, step-by-step recommendations, this article provides a comprehensive guide for educators and developers on implementing immersive, AI-supported VR courses to enhance medical education. It highlights key lessons learned in interdisciplinary collaboration, iterative testing, systematic evaluation, and alignment with educational objectives, thereby facilitating the effective, evidence-based, and sustainable integration of VR into medical curricula across diverse institutions.

Theorizing the failures of computer vision algorithms requires shifting from detecting and fixing biases towards understanding how algorithms are shaped by social, historical, and political real-world precursors. To better understand the socially embedded and historically rooted representational harms of these algorithms, we analyze how AI image captioning depicts archival images of living ethnological exibitions  (so-called 'human zoos'), mass stereotype-producing public exhibitions of colonized people common in Europe and the US from the 1870s to the 1930s, which were meant to symbolize the imagined superiority of Western societies and justify their colonial violence. We collected and analyzed more than 3800 captions from 100 archival images using MidJourney--a modern, state-of-the-art generative AI platform. Combining quantification with close reading of the captions, we found evidence of a 'colonial gaze,' an epistemological viewpoint from the perspective of colonizers characterized by significant representational harms representing five main themes: essentialism (41.6% of captions), cultural erasure (54.5%), dehumanization (11.1%), othering (28.4%), and infantilization (26.8%), with striking parallels between AI-generated captions and the original framings of human zoos informed by a broader colonial epistemology. Based on this analysis, we propose to conceptualize the colonial gaze in generative AI as an automated process of object identification and relational interpretation that draws on historical visual tropes and hierarchical logics rooted in colonial epistemologies. Trigger warning: This article contains extremely racialized text and images produced by both colonizers and the machines. The online version contains supplementary material available at 10.1007/s00146-025-02685-0.

Early detection of risk of heart failure with reduced ejection fraction remains challenging in resource-limited settings due to limited access to echocardiography. Artificial intelligence electrocardiogram (AI-ECG) algorithms have demonstrated promise for identifying left ventricular systolic dysfunction (LVSD), but their feasibility in resource-constrained settings remains unknown. To determine the frequency of patients in Kenya with a high probability of LVSD by AI-ECG and assess AI-ECG algorithm performance against the gold standard of echocardiography. This was a cross-sectional study with enrollment from June to December 2024. Participants underwent baseline assessment and 12-lead ECG, and a subset completed echocardiography within 7 days. The echocardiography subset included participants from 3 prespecified risk strata: those with prior cardiovascular disease, those at high cardiovascular risk (Framingham Risk Score [FRS] &#x2265;10%), and those at low risk (FRS <10%). The study took place at 8 outpatient health care facilities across Kenya. A total of 1444 patients 18 years and older seeking routine care were enrolled and completed paired echocardiogram. Exclusion criteria included inability to provide informed consent. Risk of LVSD was identified using a validated convolutional neural network AI-ECG algorithm (AiTiALVSD). Key outcomes were the diagnostic performance (sensitivity, specificity, and positive and negative predictive values) of the AI-ECG algorithm for detecting LVSD (LVEF <40%) when confirmed on echocardiography. Among 1444 participants (mean [SD] age, 59.0 [16.7] years; 907 [62.8%] female; 1118 [77.4%] at high risk), LVSD was identified in 204 (14.1%). The AI-ECG algorithm had a sensitivity of 95.6% (95% CI, 91.8-97.7), specificity of 79.4% (95% CI, 77.0-81.5), positive predictive value of 43.2% (95% CI, 38.7-47.9), negative predictive value of 99.1% (95% CI, 98.3-99.5), and area under the receiver operating characteristic curve (AUC) of 0.96 (95% CI, 0.95-0.97). Performance remained consistent across cardiovascular risk strata (AUC, 0.96-0.98). In this study, the AI-ECG algorithm demonstrated the potential clinical utility for screening of LVSD risk with high sensitivity and negative predictive value and may be particularly scalable in a resource-limited setting.

Basic medical pathology teaching suffers from fragmented knowledge and insufficient personalized learning, which impairs students' systematic knowledge construction and learning efficiency. This study aimed to apply an AI and knowledge graph integrated teaching model to address these challenges and provide references for basic medical education reform. An 18-week randomized controlled trial was conducted on 360 nursing students (180 each in experimental/control groups). The control group used traditional teaching, while the experimental group adopted the "AI&#x2009;+&#x2009;knowledge graph" integrated model with AI-assisted learning and knowledge graph reviews. SPSS was used for quantitative and qualitative data analysis. The experimental group showed significantly better outcomes: final score (84.2&#x2009;&#xb1;&#x2009;6.9 vs. 74.1&#x2009;&#xb1;&#x2009;7.6), after adjusting for class-level clustering using a two-level mixed-effects model, adjusted mean difference&#x2009;=&#x2009;8.2, 95% CI: 4.7-11.7, P&#x2009;=&#x2009;0.012, Cohen's d&#x2009;=&#x2009;1.13 (95% CI not computed), indicating a large effect size. knowledge mastery rate (90.3% vs.75.8%), clinical case analysis score (28.5&#xb1;3.2 vs. 21.3&#xb1;4.1, P<0.001). It also reduced ineffective learning time. This integrated model effectively promotes structured knowledge construction and learning efficiency, realizing the synergistic effect of AI and knowledge graph. It provides a replicable and practical reference for the digital and intelligent reform of basic medical education courses.

Efforts to bridge the digital divide for older adults have prioritized access and interface usability. However, the proliferation of generative AI introduces interpretive challenges that access alone cannot address. This Forum article argues that digital inclusion must evolve from operational competence to critical engagement. We introduce the concept of epistemic vulnerability to characterize a distinct risk where older adults (65+) trust AI outputs not due to cognitive deficits, but because the system's confident, fluent delivery mimics authoritative sources they have historically learned to trust. Grounding this argument in the socio-technical co-constitution model of ageing and technology, we conceptualize epistemic vulnerability as emerging from misalignments among older adults' life-worlds, design worlds, technological artefacts, and broader images of ageing. Rejecting protectionist restrictions, we propose a framework of Critical AI Literacy supported by Ethical Scaffolding, emphasizing empowerment through structural aids that help users distinguish linguistic fluency from factual accuracy. Furthermore, the risk profiles and the feasibility of scaffolding vary across the young-old (65-74), middle old (75-84) and the oldest-old (85+) subgroups. By outlining concrete strategies for adapting information-verification habits, this work provides a roadmap for ensuring older adults remain active, competent agents in an increasingly algorithmic society.

Cardiovascular diseases (CVDs) are a leading cause of morbidity and mortality globally, with low-resource settings, including Ethiopia facing challenges due to limited early diagnostic services. AI-powered electrocardiography (ECG) interpretation has the potential to improve diagnostic accuracy, decentralize care, and support timely clinical decisions, but evidence on healthcare providers' perspectives and adoption determinants is limited. This exploratory descriptive qualitative study employed 31 in-depth interviews with healthcare providers. Healthcare providers (cardiologists, internists, cardiac and critical care nurses, critical care specialists, and general practitioners) were purposively selected through maximum variation sampling from ten hospitals in four regions of Ethiopia. Data were transcribed verbatim, coded inductively, and analyzed thematically. The data analysis identified six themes: perceived benefit of AI-powered ECG interpretation CDSS, trust development, workflow integration, ethical concerns, functionality, and adoption determinants. Participants emphasized AI's potential to enhance accessibility, consistency, and diagnostic accuracy while reducing subjectivity and unnecessary referrals. Acceptance relied on high accuracy, reliable data, and rigorous validation, with the technology seen as supportive rather than replacing clinicians. Material resources, human resource readiness, and leadership engagement were key factors for adoption. Recommendations included phased implementation, continuous training, and model expansion to ensure sustainability and clinical utility. The AI-powered ECG interpretation CDSS was viewed as a valuable adjunct for strengthening cardiovascular care in Ethiopia, highlighting the need for context-sensitive strategies, ethical safeguards, and multi-level system readiness for successful adoption.

The integration of Artificial Intelligence (AI) into medicine has progressed from discriminative models to Generative AI (GenAI), which can synthesize novel content. For orthopaedic surgeons, scientific publication remains a vital marker of academic success but is often constrained by clinical workload. This review proposes a structured, practical framework to help orthopaedists effectively harness AI tools, transitioning from opaque, "black box" generation to grounded, verifiable research assistance through Retrieval-Augmented Generation (RAG). A PubMed search was conducted to explore the application of GenAI in the context of orthopaedic scientific research. An interactive review with experts in GenAI was also conducted, from which the proposed structure was developed. From this synthesis, a three-phase workflow is proposed: (1) Evidence selection using semantic discovery systems to identify and map relevant literature beyond keyword matching; (2) Data extraction and synthesis employing RAG-based systems to anchor AI responses to verified PDF sources, thereby minimizing hallucinations; and (3) Drafting and refining using Large Language Models (LLMs) for structured composition, linguistic clarity, and iterative manuscript improvement. The workflow integrates platform features to enhance efficiency, accuracy, and accessibility in orthopaedic research. When applied within a controlled, evidence-grounded environment, these systems can automate literature synthesis, expedite data extraction, and assist with scientific writing, while preserving authorial intent and accountability. However, challenges remain. Risks include algorithmic bias, "hallucinations", privacy concerns, and ethical issues related to authorship. Despite these limitations, AI represents a paradigm shift in orthopaedic scholarship, functioning as a cognitive exoskeleton that augments rather than replaces human expertise. With vigilant human oversight and adherence to journal ethics, orthopaedic surgeons can leverage AI to enhance research productivity, reproducibility, and quality while upholding the highest standards of scientific integrity.

To assess the cost-effectiveness of using artificial intelligence (AI)-derived software to assist reading CT scans of the chest to identify and analyse lung nodules compared to unaided reading in symptomatic, incidental and screening populations. Decision tree structures were developed in TreeAge Pro 2021. Structures were informed by British Thoracic Society clinical guidelines and clinical opinion. Results were presented as incremental cost-effectiveness ratios (ICERs) expressed as cost per quality-adjusted life-year (QALY) over a lifetime from the UK National Health Service and Personal Social Services perspective. For the symptomatic population, the unaided radiologist reading strategy dominated the AI-assisted reading strategy. In the incidental population, unaided radiologist reading was cost-effective with an ICER of approximately &#xa3;1000 per QALY. Conversely, in the screening population, AI-assisted radiologist reading dominated unaided reading. The cause of AI assistance being cost-effective depended on the number of people who had undergone CT surveillance because of non-cancerous findings. Given the limitations in the quality and quantity of evidence to inform inputs, these results should be interpreted with caution. Current analyses based on limited evidence suggested that, in the symptomatic and incidental populations, unaided radiologist reading may be the more cost-effective strategy, while in the screening population, AI-assisted radiologist reading appeared to be the dominant strategy. Better quality evidence is required to have a definitive answer about their cost-effectiveness. This paper shows whether adding AI-derived software to radiologists' reading of CT scans to identify lung nodules offers good value for money.

Multimodal artificial intelligence systems combining text and image analysis represent a paradigm shift in clinical decision support. While GPT-4 with Vision (GPT-4V) has shown promise in medical imaging interpretation, existing studies report inconsistent performance (16%-80% accuracy) across radiological subspecialties. Critical knowledge gaps persist regarding GPT-4V's capability to integrate clinical history with imaging findings in complex neuroradiology scenarios, and fundamental questions remain about whether the model appropriately balances visual and textual information sources when formulating diagnoses. Furthermore, documented artificial intelligence hallucination rates of 35.5% to 63% in radiology applications raise urgent safety concerns, yet the relationship between modality utilization patterns and diagnostic accuracy remains unexplored. This study aims to evaluate GPT-4V's diagnostic accuracy on expert-validated neuroradiology board-style examination questions and to examine the model's self-reported reliance on imaging versus clinical text data when making diagnostic decisions. A secondary objective was to examine whether self-characterized modality utilization patterns differed systematically between correct and incorrect diagnoses, potentially identifying specific failure modes requiring targeted mitigation strategies. This cross-sectional study evaluated GPT-4V using 29 neuroradiology cases from the RSNA (Radiological Society of North America) Case Collection, covering adult brain and central nervous system pathologies imaged via computed tomography or magnetic resonance imaging. The cases were authored by board-certified radiologists. GPT-4V was accessed via ChatGPT Plus (July 2024) with standardized prompts selecting 1 answer from 4 options, providing diagnostic rationale, and quantifying the percentage contributions of image versus text data. Binary scoring assessed diagnostic performance (correct=1, incorrect=0). Statistical analysis included Wilson score CIs, a binomial test comparing accuracy to chance, and a 2-tailed t test comparing self-reported modality reliance between correct and incorrect diagnoses (&#x3b1;=.05, Cohen d calculated). GPT-4V correctly diagnosed 22 of 29 cases (76% accuracy, 95% CI 57.9%-87.8%), significantly exceeding the chance performance of 25% (z=6.33; P<.001). The model self-reported mean contributions of 66.1% from imaging (95% CI 63.5%-68.8%) and 33.9% from text (95% CI 31.2%-36.5%). Correct diagnoses (n=22) showed significantly lower self-reported image reliance (62.8%, 95% CI 61.3%-64.3%) compared to incorrect diagnoses (n=7; 76.7%, 95% CI 73.5%-80.0%), with a mean difference of 13.9 percentage points (95% CI 10.6-17.3; P<.001; Cohen d=4.08, 95% CI 2.73-5.43). All 7 incorrect diagnoses demonstrated image-dominant attribution &#x2265;70% (Fisher exact test P<.001), suggesting that excessive visual reliance may indicate diagnostic risk. The 76% accuracy substantially exceeds prior GPT-4V radiology studies (43%), demonstrating that focused domain application with structured prompting enhances performance. Incorrect diagnoses are associated with higher self-reported visual reliance, suggesting a potential failure mode warranting experimental validation. This pattern identifies a potentially actionable signal for quality assurance systems. Clinical deployment should remain restricted to supervised educational applications with mandatory radiologist oversight until balanced context-aware integration is validated.

Background: In an earlier murine model of myocardial infarction (MI), we showed that CD8 cells and myeloid dendritic cells (mDCs) infiltrate the infarcted myocardium within the first week. However, in humans, the spatial interplay between CD8+ T cells and dendritic cells in the spatial context of human myocardial infarction remains underexplored. Objective: In the present study, we applied spatial transcriptomics and functional assays to characterize immune-stromal dynamics in infarcted myocardium and peripheral blood. Methods & Results: Spatial transcriptomics analysis of infarcted human myocardium at days 2 and 6 post-MI, combined with peripheral blood flow cytometry and EPC colony-forming assays, was performed. Cell composition, pathway enrichment, and cell-to-cell communication analyses were conducted to map immune-stromal cells' dynamics across time points. Spatial mapping identified dynamic shifts in immune, fibroblast, and endothelial populations, with fibroblasts and endothelial cells remaining abundant throughout. CD8+ T cells accumulated in ischemic regions while their circulating levels declined. Gene Ontology and pathway analyses of CD8A+ transcripts revealed enrichment of proinflammatory and NF-&#x3ba;B survival programs. ITGAX/CD33/THBD+ APCs progressively increased within infarct zones, activating antigen-presentation and leukocyte chemotaxis pathways. Early (day 2) APC-endothelial crosstalk showed the strongest predicted recruitment signals for CD8+ T cells, which diminished by day 6. Finally, EPC colony-forming capacity showed a tendency for reduction in MI patients and inversely correlated with coronary lesion burden, indicating impaired vascular repair potential. Conclusions: This integrative spatial and functional study demonstrates that APC-driven CD8+ recruitment and EPC dysfunction are key features of human MI. Immune-endothelial niches facilitate early cytotoxic T-cell infiltration, while progenitor depletion limits vascular regeneration. These findings provide mechanistic insight into immune-vascular imbalance during infarct healing and highlight potential therapeutic targets to modulate inflammation and restore vascular repair.