Psyche and pixels: AI-generated images and their impact on stigma of mental illnesses in Germany.

Few-shot learning for surgical phase recognition: Performance and generalization in cholecystectomy.

Disentangled generative uncertainty-aware multi-modal diffusion segmentation of medical images.

The global expansion of English Medium Instruction (EMI) in medical education has intensified demands on students who must simultaneously acquire disciplinary knowledge and operate in a non-native language. This study investigates the coping strategies employed by Arabic-speaking medical students across three Saudi public universities, contributing empirical evidence on how learners adapt to linguistically complex EMI medical contexts. A convergent mixed-methods design combined semi-structured interviews (n = 50 female students), questionnaires (n = 289), and non-participant observations (n = 9) across classrooms, study groups, and laboratories. Qualitative data underwent thematic analysis using NVivo; quantitative data were analysed using descriptive statistics (frequencies and percentages) in SPSS version 27. Triangulation enhanced validity, with trustworthiness ensured via member checking and reflexivity. Students deployed a dynamic repertoire of strategies: multimodal resource use (e.g. 60% commonly watched Arabic YouTube videos, 25% English), translanguaging and translation (68% translated terms), memorisation via flashcards, and collaborative practices (e.g., group translation of slides). Observations confirmed these as central mechanisms for accessing, processing, and consolidating medical knowledge, evolving with proficiency. These practices are essential for disciplinary learning in EMI medical education and can also be understood as preparing students for participation in multilingual medical settings where English functions as a shared professional resource. Findings offer insights for policy, curriculum design (e.g., legitimising translanguaging, embedding multimodality), and pedagogical and language support, particularly where content is taught through a foreign language. They underscore the need for bilingual approaches to enhance clinical readiness and equity in EMI medical contexts.

The European Union's Artificial Intelligence Act introduces a novel regulatory category of "unacceptable risk," prohibiting specific AI practices that are deemed fundamentally incompatible with human rights and ethical principles. While much attention has focused on the regulation of high-risk AI systems, particularly in medical contexts, the AI Act's outright bans under Article 5 have received far less scrutiny. This paper addresses that gap by examining how these prohibitions apply to healthcare and public health, which are domains defined by rapid technological uptake, structural vulnerability, and ethically sensitive decision-making. Drawing on the European Commission's 2025 interpretative Guidelines, the paper argues that several health-related AI applications, such as emotion recognition tools, biometric categorisation systems, and technologies that influence or target vulnerable populations, may fall within the scope of the bans. It also shows that the Act's medical and safety exceptions risk weakening the vulnerability protections that the prohibitions aim to secure. By integrating legal analysis with real-world health examples, the paper offers a framework for interpreting these prohibitions and assesses how they should guide the ethical boundaries of AI in healthcare, within and beyond the European context.

Early pregnancy is commonly accompanied by a wide range of symptoms, yet psychological determinants of pregnancy-related symptom burden remain poorly understood. Evidence from other medical contexts suggests that expectations may contribute to symptom burden. This prospective study examined whether first-trimester symptom expectations predict pregnancy-related symptom burden and disability at the beginning of the second trimester. In the PrExpect study, pregnant women up to the 10th gestational week completed online questionnaires assessing pregnancy-related symptoms, symptom expectations, psychological variables, and pregnancy characteristics, with follow-up at the beginning of the second trimester. Hierarchical regression models identified predictors of symptom burden and disability. Bidirectional temporal associations between symptoms and expectations were examined using a cross-lagged panel model. Exploratory analyses assessed pregnancy-related mindsets. A total of 347 women provided baseline data, and 227 completed follow-ups. Symptom burden was high and increased across the first trimester. Stronger baseline symptom expectations consistently predicted greater symptom burden and disability at follow-up. Cross-lagged analyses indicated bidirectional associations between expectations and symptoms over time. Several negative pregnancy-related mindsets were associated with higher symptom burden. These findings identify symptom expectations as a psychological correlate of early pregnancy-related symptom burden and disability, suggesting that symptom expectations may represent a promising target for future intervention studies to support well-being in early pregnancy.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-57961-w.

ABSTRACT Introduction Digital technologies are increasingly incorporated into dental education to improve training and promote more objective skill assessment. Digital superimposition enables quantitative comparison between student preparations and reference models but is seldom applied in crown preparation training. Conventional methods using extracted teeth lack standardization and provide limited quantitative feedback. This study evaluated whether a digital superimposition–based assessment framework using standardized 3D‐printed teeth yields objective measurements for expert‐defined pass–fail evaluations of crown preparations. Materials and Methods Thirty‐three dental interns prepared crowns on standardized 3D‐printed mandibular molar models. The preparations were scanned and digitally superimposed onto a predefined standard to calculate volumetric and surface area deviations. A single‐group pre–post design was adopted to ensure educational equity in irreversible skill training. Pass–fail outcomes were determined through combined clinical and technical evaluations. Changes between pre‐training (T1) and post‐training (T2) assessments were analysed using McNemar's exact test and geometric discrepancies were examined for construct‐related validity. After T2, the participants were surveyed for their perceptions. Results The proportion of acceptable preparations increased significantly from 30.3% at T1 to 69.7% at T2, with no performance regression. Acceptable preparations showed significantly smaller volumetric and surface area deviations from the digital standard compared to non‐acceptable preparations. The participants perceived high value in self‐evaluation and interprofessional communication. Conclusion Digital superimposition analysis using standardized 3D‐printed models provides objective metrics of expert‐defined preparation quality. This framework facilitates formative assessment within a competency‐based medical education context and enhances interprofessional education by bridging the gap between clinical and technical standards.

Minimizing nocebo side effects in health care: The unrecognized role of patient involvement.

Technology-enhanced learning is increasingly applied in medical education to support complex cognitive processes, including diagnostic reasoning. Point-of-care ultrasound (POCUS), as a portable and real-time imaging modality, has been widely integrated into clinical training. However, limited research has explored how learners engage with POCUS to construct diagnostic understanding, particularly within international medical education contexts involving students from resource-variable healthcare environments. This study aimed to investigate international medical students' learning experiences with POCUS as a technology-enhanced learning tool. A qualitative study was conducted involving 28 third- and fourth-year international medical students who participated in a structured POCUS learning activity. The intervention combined guided instruction with supervised hands-on practice in cardiac and abdominal ultrasound within an emergency medicine training context. Following the activity, participants completed written reflective narratives. Data were analysed using reflexive thematic analysis to identify recurring patterns in learners' experiences and meaning-making processes. Four major themes were identified. First, POCUS facilitated diagnostic understanding through real-time visualisation of anatomical structures and physiological processes. Second, integration of ultrasound within problem-based clinical scenarios supported the development of diagnostic reasoning. Third, collaborative learning and instructor scaffolding enhanced engagement and knowledge construction. Fourth, participants recognised the limitations of POCUS and highlighted the importance of reflective practice and continued skill development. These findings demonstrate that learners actively integrate visual, experiential, and contextual information when engaging with POCUS. POCUS functions as an effective technology-enhanced learning tool that supports diagnostic reasoning by linking visualisation with contextualised and reflective learning processes. The findings extend current understanding of technology-enhanced medical education by demonstrating how domain-specific technologies facilitate meaning-making and cognitive integration in diverse and cross-cultural learning environments. These insights have implications for the design of ultrasound-based curricula in emergency medicine education.

Clinical Characteristics and Treatment of Benzodiazepine Withdrawal Delirium in Adults: An Illustrative Case and Systematic Review of Published Cases.

BackgroundPediatric heart disease (PHD), including congenital heart defects, is often incompletely captured in electronic health records, particularly when clinical significance must be inferred from unstructured echocardiogram reports. Automated methods capable of extracting clinically meaningful PHD from narrative reports could improve clinical decision support and research applications.ObjectiveThe aim of the study is to evaluate the feasibility of using supervised fine-tuning of large language models (LLMs), with and without chain-of-thought (CoT) reasoning, to characterize patients with clinically significant or historical PHD from unstructured echocardiogram reports.MethodsWe developed a PHD detection algorithm using fine-tuned open-source LLMs, including LLaMA (Meta) and Qwen (Alibaba), to analyze 9749 echocardiogram reports. A subset of 712 reports was adjudicated by 2 pediatric cardiac anesthesiologists, classifying 506 (71.1%) as clinically significant PHD and 206 (28.9%) as not significant. While DeepSeek R1 has shown improved performance with CoT reasoning, its application in medical contexts is underexplored. We incorporated R1-generated CoT into model prompts and fine-tuned backbone LLMs.ResultsThe fine-tuned Qwen-7B-10k-overthink-CoT achieved the highest accuracy (92.4%), outperforming Qwen-7B-without-CoT (90%), LLaMA-3B-without-CoT (87.9%), Qwen-3B-without-CoT (85.6%), Qwen-3B-10k-overthink-CoT (68.5%), and LLaMA-3B-10k-overthink-CoT (46.2%). In a second dataset, an external validation was performed (n=113; 64 positive, 49 negative), Qwen-7B-10k-overthink-CoT sustained a strong, balanced performance (82.7%), followed by Qwen-7B-without-CoT (88.4%), LLaMA-3B-without-CoT (86.8%), Qwen-3B-without-CoT (84.5%), Qwen-3B-10k-overthink-CoT (58.9%), and LLaMA-3B-10k-overthink-CoT (46.2%). The fine-tuned Qwen-7B model with overthinking CoT (10,000 tokens) achieved the highest internal accuracy (92.4%), with balanced sensitivity and specificity. Across repeated runs, CoT-enhanced models demonstrated improved classification consistency compared to non-CoT models (Qwen-7B-without-CoT: 90%, LLaMA-3B-without-CoT: 87.9%, Qwen-3B-without-CoT: 85.6%). In external validation (n=113), non-CoT variants achieved higher accuracy (up to 88.4%), whereas the Qwen-7B CoT model demonstrated more balanced class performance (accuracy=82.7%).ConclusionsSupervised fine-tuning of LLMs with CoT offers an effective approach for automated PHD detection within unstructured data in the electronic medical record. While CoT-enhanced models demonstrated improved internal performance and more balanced classification, they did not consistently achieve higher accuracy in external validation, highlighting trade-offs between accuracy and class balance. These findings highlight the promise of LLM-based approaches for clinical text phenotyping while underscoring the need for larger, multicenter validation and careful calibration for real-world deployment. Continued validation and integration into the electronic medical record are essential for real-world, artificial intelligence–driven clinical decision support.

This article examines the intersection of anxiety, emotional geography, and bioethics, exploring how uncertainty shapes the experience of patients navigating medical and social landscapes. When is the good time to start worrying about one's own well-being? Who is the "good patient"? Anxiety, particularly in medical contexts, arises from the epistemic tension between known and unknown realities, challenging patients and clinicians alike to discern which concerns warrant attention. Drawing on phenomenological and bioethical perspectives-including those of Heidegger (1962), Sartre (1996), Ricoeur (1992, 2004), and Tillich (1952), as well as the phenomenology of the body developed by Leder (1990)-alongside contemporary scholarship in emotional geography and medical humanities, this paper investigates different forms of anxiety: digital, existential, and spatial, and their impact on subjective and intersubjective medical experiences. Special attention is given to the "dys-appearing" body, the disruption of embodied self-evidence in illness, the courage required to exist in the face of medical uncertainty, and the constitutive role of intersubjective environments in shaping patients' sense of well-being. Ultimately, I argue for a more compassionate, intentional engagement with anxiety that transcends mere symptom management, advocating instead for a holistic approach that accounts for the emotional, spatial, and existential constitution of patients' realities.

This article introduces a biocompatible, non-invasive intra-body communication (IBC) system that employs a U-shaped, flexible wearable antenna specifically designed for fat-intra-body communication (Fat-IBC) applications. The proposed Fat-IBC system is engineered to function within the industrial, scientific, and medical (ISM) frequency range of 2.40–2.48 GHz. The U-shaped wearable antenna is encased in a biocompatible polydimethylsiloxane (PDMS) coating layer, enhancing durability and safety, making it suitable for continuous use in medical contexts. The U-shaped Fat-IBC antenna leverages the low-loss property of adipose (fat) tissue, positioned between skin and muscle layers, to facilitate efficient signal propagation across the body. The antenna is optimized for transmission through a three-layer tissue model (skin, fat, and muscle), leveraging the favorable transmission properties of adipose tissue to minimize signal loss. To assess performance, detailed numerical simulations and experimental validations were conducted using three-layer tissue models, torso phantoms (obese and athletic), and human volunteer trials. Ethical approvals were obtained for human testing, ensuring compliance with biomedical research standards. The performance of the U-shaped Fat-IBC antenna was benchmarked against a standard Bluetooth low energy (BLE) chip antenna under controlled conditions, with measurements taken in both front-to-back and side-to-side orientations across various body types. Experimental results demonstrated the superior performance of the U-shaped Fat-IBC antenna over BLE, specifically in terms of signal stability and transmission efficiency across body tissue, especially within adipose layers. Specific Absorption Rate (SAR) analysis revealed a peak value of 0.3061 W/kg, which is well within the IEEE safety limits, confirming the system’s suitability for biomedical applications. This Fat-IBC platform demonstrates the feasibility of fat tissue as a viable medium for intra-body networks, offering a reliable, power-efficient solution for real-time medical telemetry and patient monitoring, with promising implications for broader health applications.

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) remains the only curative treatment for many hematologic malignancies, yet patient outcomes vary significantly. Patient expectations influence recovery in other medical contexts, yet their role in allo-HSCT remains unclear. This pilot study examined whether pre-transplant treatment expectations predict psychological and immunological outcomes post-transplant. In this prospective, single-center observational cohort study, 42 patients undergoing allo-HSCT were assessed at baseline (T0), discharge (T2), and six months post-transplant (T3). Questionnaires measured illness-related disability (PDI, primary endpoint at T3), treatment expectations (TEX-Q), quality of life (FACT-Leu), depression (PHQ-9), and anxiety (GAD-7). Immunological markers, including inflammatory markers were collected at T0 and T3. Baseline-adjusted regression analyses with full-information maximum likelihood estimation were used. P-values were corrected for multiple comparisons using a false discovery rate approach. Baseline expectations were associated with psychological outcomes at hospital discharge and immunological and inflammatory markers at six-month follow-up: For instance, negative impact expectations were associated with higher disability (β=0.522, p<0.001), depression (β=0.693, p=0.009), anxiety (β=0.737, p=0.003), and lower quality of life (β=-0.576, p<0.001) at T2. Benefit expectations were associated with higher lymphocyte counts (β=0.453, p<0.001) and lower CRP levels at T3 (β=-0.28, p=0.011). Positive impact expectations were associated with more favorable T-cell subsets. Pre-transplant expectations may influence psychological and immune recovery following allo-HSCT. Addressing expectations could enhance outcomes and should be explored in future intervention studies.

In the 19th and early 20th centuries, many fetuses and infants were collected for anatomical study. Yet little research has explored their origins or the ethical implications of holding and using these individuals in teaching and research. This paper reviews the literature on fetal and infant skeletal collections and presents a case study from the W. D. Trotter Anatomy Museum in Aotearoa New Zealand that uses the lens of biopower and necropolitics to examine the acquisition and use of these individuals in anatomical education. This model highlights how societal and institutional powers and medicine regulate reproductive bodies and determine the perceived worth of fetuses and infants within specific cultural, historical, and medical contexts. This case study shows that individuals were often acquired from marginalized populations-unwed parents, the impoverished, and institutionalized women-whose reproductive autonomy was politically and socially negated. The paper explores how biopower and the necropolitics of reproduction in colonial New Zealand operated to control populations and individuals. Biopower and the necropolitics of reproduction were enacted through eugenic sentiment, structural inequality in healthcare, alongside medical and institutional control over the living and the dead. This contributed to higher infant mortality for marginalized mothers and infants, a loss of autonomy over the fate of deceased bodies, and the suppression of grief. The anatomization of fetuses rendered them objects of scientific value, whilst simultaneously erasing their personhood and socio-historical context, thus extending the structural violence their families experienced during life into their postmortem "life" history (necro-violence).

Difficult but doable: Health professional perspectives of assessing clinical deterioration in children with dark-coloured skin.

Neurological emergencies are among the most time-sensitive and high-stakes presentations in the emergency department, yet they often remain underrepresented in traditional training curricula. To address this gap, the Neurological Emergencies Card Game (NExCARD) was developed as a gamified, card-based educational intervention aimed at enhancing emergency medicine residents' competence and confidence in diagnosing and managing neurological conditions. Inspired by the mechanics of the card game UNO, NExCARD integrates disease components into colour-coded, clue-based cards played competitively in team-based sessions. Post-activity feedback from 27 residents revealed increased engagement, a deeper understanding of the management of neurological emergencies, and a strong preference for incorporating similar tools into regular academic sessions. This commentary explores the rationale, structure, implementation, and reception of NExCARD, reflects on the educational theories that support gamification, and proposes ways in which it may be adapted, scaled, and evaluated in broader medical education contexts. Key Words: Neurology, Emergency medicine, Gamification, Learning, Skills.

To identify and describe components of the home-based palliative care model for patients with heart failure, re-conceptualize the process, and update the existing model to inform nursing research and clinical practice. Heart failure, as a chronic progressive disease, has a continuously increasing incidence and medical burden, and requires multidimensional clinical management strategies. Although palliative care is recommended for improving the quality of life of heart failure patients, its utilization rate is still low, and the core components and implementation pathways of home-based palliative care are not yet clear. Using Rodgers' Evolutionary method, a concept analysis was conducted to define the core components for home-based palliative care in heart failure. Data were drawn from professional literature spanning 1990 to 2025, using the terms "home-based palliative care," "home palliative care," "home palliative treatment," "family hospice care," and "heart failure," "cardiac insufficiency," "congestive heart failure". Abstracts from 101 articles were initially reviewed, with 22 articles retained for analysis. Core concepts were identified, defined, and synthesized. The PRISMA 2020 checklist was used. The refined model clarifies nurses' roles in linking home-based palliative care core attributes: self-care, collaborative palliative care, communication of care goals, and support. It presents patients', family caregivers', and healthcare workers' impact on home-based palliative care triggering, implementation, and outcomes, emphasizing patient-centered continuous evaluation/adjustment. This model provides a structured framework to improve heart failure home care delivery and enhance patient/family quality of life, with potential to inform home palliative care approaches for other chronic illnesses. Further refinement/adaptation is needed to ensure relevance across medical contexts.

Characterization and measurement of reflective capacity (RC) - the ability to learn from experience to inform future learning and practice-in hospital trainees are critical to the design of medical education programs; testing interventions requires effective assessment. This study assessed RC among anesthesia trainees using the Reflection Evaluation for Learners' Enhanced Competencies Tool (REFLECT), which evaluates reflective writing. The study explored variation by training level, age, program, gender, and examined reliability and validity against an independent reflection measure. A prospective observational study (February-May 2021) included anesthesia trainees across four training levels, four programs, and varied prior RC experience and intended sub-specialties. Participants completed the REFLECT instrument (using simulated and personal events) and the Residency Level Cognitive Behavior Survey (rCBS). Potential influencing factors (scenario realism, perceived importance, prior reflective writing) were examined. Data were analyzed for internal consistency (t-tests, ANOVA, Pearson's r) and interrater reliability (Cronbach's α, intraclass correlation coefficient). Participants (n = 172) demonstrated moderate RC (mean = 2.5/4, SD = 0.45), with no significant differences by training level, program, gender, or age. Greater perceived realism and prior reflective writing were independently associated with higher RC. Those with prior RC training rated reflection as more important and useful. Personal and simulated event scores were similar. REFLECT showed excellent interrater reliability (Cronbach's α = 0.95) and moderate correlation with the rCBS Reflection domain (r = 0.31, P = .001). REFLECT demonstrates robust psychometric properties for assessing RC in anesthesia trainees. Moderate RC scores support the need for interventions to enhance reflection. Although limited to one specialty and not assessing application to practice, the instrument's general structure suggests potential utility across postgraduate medical training contexts.

BackgroundBullying of medical students by clinicians is a common problem in medical education that has adverse effects on student psychological well-being, educational experiences and professional development. The hierarchical structure, power imbalances and historical teaching practices present in South African medical schools may contribute to an environment where that mistreatment occurs, but limited synthesis of the evidence exists in relation to South Africa.AimThe aim of this scoping review is to review and map existing literature related to the bullying of medical students by clinicians in South African medical training contexts, including the forms, prevalence, contributing factors, consequences and how institutions respond.Methods and analysisThe authors will execute this scoping review in conformity with the Joanna Briggs Institute manual for evidence synthesis and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Review (PRISMA-ScR) guidelines. This will be a mixed-method scoping review that incorporates both qualitative and quantitative studies. The review process and procedures will include a search in databases such as PubMed, Cumulative Index to Nursing & Allied Health Literature, Scopus, Web of Science, Psych-Info and African Journals Online (AJOL). Socioecological model is a model that incorporates multiple behaviour change across five domains. The review is expected to take 10 months to complete, commencing on 10 November 2025 and concluding on 10 September 2026. Thematic analysis will be conducted with NVivo data management software.PROSPERO registration numberCRD420251179182.