人工智能

In periodontology, Artificial Intelligence (AI) applications, ranging from radiographic evaluation to outcome prediction, are emerging. However, their adoption is impacted by practitioners' awareness, attitudes, and perceived barriers. Evidence regarding AI adoption among Egyptian periodontists remains limited. Therefore, this study aimed to assess knowledge, perceptions, usage, and concerns regarding AI among Egyptian periodontists, and to identify demographic and professional factors associated with AI adoption. A multicenter cross-sectional online survey using a 33-item questionnaire was uploaded via Google Forms and distributed to eligible Egyptian periodontists. To gather information about participants' knowledge, opinions, and concerns about artificial intelligence in periodontology, the survey used closed-ended questions with a 3-point Likert-type scale. A total of 275 Egyptian periodontists took part. Although familiarity with AI was high (98.2%), only 31.3% reported understanding its working principles. Attitudes toward AI were generally positive, with 89.8% considering it a new era, and 80.7% expecting it to significantly improve periodontology. Despite substantial interest, practical familiarity with AI-based dental software remained limited (10.9%), with research being the most common application area (58.2%), followed by implant planning (13.1%) and diagnosis (12.7%). The main concerns centered on over-reliance on AI affecting critical thinking skills (68.4%), the reliability of AI-assisted periodontal diagnosis (65.8%), security risks (59.6%), and patient privacy issues (53.1%). Ordinal logistic regression analyses identified several factors significantly associated with AI-related outcomes, with AI working principle knowledge associated with age and professional experience. Male gender was significantly associated with the perception that AI could replace periodontists. AI-related educational engagement was associated with age, professional experience, and institutional affiliation. AI-related concerns were associated with gender and educational level. The findings suggest a gap between high awareness and limited adoption of AI among Egyptian periodontists despite generally positive attitudes. Key demographic and professional variables such as age, experience, gender, and institutional affiliation emerged as significant associated factors in the regression models for knowledge, perception, usage, and concerns. The study results highlight potential educational gaps and support the integration of AI education in postgraduate programs.

Artificial intelligence (AI) is rapidly transforming healthcare, with growing impact on maternal and child health (MCH) through advances in machine learning, deep learning, computer vision, generative models, and conversational systems. This article provides a comprehensive synthesis of current AI applications in MCH, structured across six key domains: predictive modeling, image analysis, deep learning and interpretability, generative and multi-omics approaches, conversational AI, and environmental and lifestyle analytics. Drawing on recent literature and the translational experience of the Spanish RICORS-SAMID network, we analyze how these technologies are being integrated into clinical, preventive, and assistive workflows. Across domains, AI demonstrates strong potential for early risk prediction (e.g., preeclampsia, fetal growth restriction, neonatal outcomes), automated image interpretation, biomarker discovery, and personalized decision support. However, despite promising performance metrics, most systems remain at the proof-of-concept stage, with limited external validation, scarce prospective evaluation, and incomplete integration into real-world clinical pathways. Key translational gaps include data heterogeneity, lack of interoperability, insufficient explainability, and challenges related to bias, fairness, and regulatory compliance. We argue that the next phase of AI in MCH must shift from static prediction toward longitudinal, mechanism-aware, and clinically actionable systems, supported by robust validation and multidisciplinary collaboration. Particular emphasis is placed on equity, as the benefits of AI must extend to low-resource settings where maternal and neonatal morbidity remains highest. By bridging technical innovation with clinical implementation, coordinated research networks such as RICORS can play a critical role in accelerating the safe, effective, and equitable deployment of AI in maternal and child healthcare. Artificial intelligence (AI) refers to computer systems that can learn from data and help identify patterns, make predictions, or support decision-making. In maternal and child health, AI is increasingly being used to help healthcare professionals detect complications earlier, improve diagnoses, and provide more personalized care for mothers, babies, and children. This article reviews how AI is currently being applied across different areas of maternal and child health. Examples include identifying women at risk of conditions such as preeclampsia or gestational diabetes, predicting premature birth or newborn health problems, analyzing medical images, supporting mental health assessment, and studying how environmental factors may affect pregnancy and child development. AI is also being used to combine different types of health information, which may help doctors better understand complex diseases and tailor care to individual patients. Although many AI systems have shown promising results in research studies, only a small number have been fully tested in routine healthcare settings. Before these technologies can be widely adopted, researchers and healthcare organizations must ensure that they are safe, reliable, fair, and effective for diverse populations. It is also important to maintain patient trust, protect privacy, and ensure that AI supports rather than replaces healthcare professionals. The future of AI in maternal and child health is likely to involve systems that continuously learn from multiple sources of information and help provide more proactive and personalized care. However, achieving these benefits will require collaboration between clinicians, researchers, engineers, patients, and policymakers. Ultimately, the goal of AI is not to replace human expertise, but to help healthcare teams deliver better care and improve outcomes for mothers, children, and families.

Artificial intelligence (AI) is increasingly reshaping medical and health-professions education through adaptive tutoring, generative content creation, simulation analytics, automated assessment, and diagnostic-reasoning support. Since 2023, large language models and multimodal AI systems have expanded AI from relatively narrow analytic tools into interactive educational agents capable of dialogue, feedback, and content generation. This narrative review asks: How does AI contribute to medical education across four core educational functions, and what methodological, ethical, and implementation constraints should guide responsible integration? To answer this question, we synthesize evidence across four functional domains: AI-supported instruction and content generation, AI-augmented simulation and procedural training, AI-supported assessment and learning analytics, and AI-assisted diagnostic reasoning and clinical cognition. A structured search of PubMed, Scopus, and Web of Science (January 2018-January 2025), supplemented by citation tracking, identified 540 records, of which 90 met the eligibility criteria. Evidence was analyzed using an iterative qualitative synthesis approach that combined inductive coding of educational mechanisms with deductive organization into the four predefined functional domains. Findings were summarized descriptively because of substantial heterogeneity in interventions, outcomes, and study designs. In instruction and content generation, AI-supported tools were associated with preliminary gains in knowledge acquisition, learner engagement, and self-regulated study, although accuracy depended on supervision and prompt quality. In simulation and procedural training, computer-vision, motion-analytics, and VR/AR systems supported immediate feedback and were associated with improved procedural efficiency in early studies. In assessment and learning analytics, AI tools reduced feedback latency and improved scoring consistency, but validity, fairness, and explainability remained central concerns. In diagnostic reasoning, AI-supported case platforms and LLM-based dialogue improved short-term reasoning performance and metacognitive calibration, while evidence for long-term clinical transfer remained limited. AI appears most defensible as an augmentative educational partner that strengthens feedback, personalization, and competency-based progression, rather than as an autonomous substitute for educators. The current evidence supports cautious, supervised implementation, but remains early-phase and methodologically uneven. Multicenter validation, transparent reporting, equity-focused implementation, and structured AI-literacy training are essential before AI can be integrated more broadly into high-stakes educational workflows.

Patients with muscle-invasive bladder cancer (MIBC) have heterogeneous outcomes following transurethral resection of bladder tumor (TURBT). We used a computational histopathology artificial intelligence (CHAI)-based platform to develop and validate a digital image-only MIBC prognostic biomarker. The CHAI platform extracts histologic features from pre-treatment TURBT specimen H&E-stained whole slide images. The Cancer Genome Atlas was used for development to construct a signature of features associated with the primary endpoint of recurrence-free survival (RFS). A continuous risk score was dichotomized into favorable and unfavorable groups. For validation, the performance of the locked model was then assessed in an independent, held-out, retrospective, pooled real-world data cohort of patients from NCI-Designated Cancer Centers with cT2N0M0 urothelial carcinoma who underwent radical cystectomy with/without neoadjuvant chemotherapy (NAC). A total of 178 patients were included: 44 in development and 134 in validation, of whom 50% received NAC. In validation, those classified as unfavorable risk by the CHAI biomarker (N = 67) had worse RFS (HR 3.1 [1.7-5.7], P < 0.001), cancer-specific survival (CSS) (3.5 [1.5-7.8], P = 0.003), and overall survival (OS) (3.0, [1.5-5.7], P = 0.001) vs. favorable risk (N = 67). Three-year RFS was 40% vs. 74% for disease classified as unfavorable and favorable risk, respectively (P < 0.001). After adjusting for prognostic clinical variables, including receipt of NAC, the biomarker remained associated with RFS, CSS, and OS (P < 0.01). Exploratory analysis found a significant interaction between the biomarker and NAC for RFS (P = 0.02). We developed and validated an image-only AI-based biomarker from pre-treatment H&E TURBT specimens associated with clinical outcomes in cT2 MIBC. While future development and validation work is warranted, these hypothesis-generating retrospective findings support the potential of this approach to advancing precision medicine in MIBC.

The expansion of artificial intelligence (AI) in healthcare is often framed as a response to clinical needs and system inefficiencies; however, its alignment with professional values and everyday practice remains less examined. This study examines the role of AI in dentistry through qualitative semi-structured interviews with 16 dentists in active clinical practice, conducted in Chile in 2025, focusing on how practitioners evaluate AI in relation to their professional ethos. The findings show that dentistry is understood as a moral practice centered on relational care and situated judgment. While technology has historically supported this ethos, the processes of marketization, bureaucratization, and acceleration increasingly undermine its conditions. AI is not simply experienced as being beneficial or harmful. Applications that reduce administrative burden or support diagnostic tasks are often welcomed by practitioners, and in isolation, this assessment is reasonable. However, when examined at the level of the sociotechnical system in which they are embedded, even tools that appear to return time to the clinical encounter operate within institutional arrangements that simultaneously compress time and intensify productivity demands. The study further argues that AI development does not consistently follow articulated clinical needs but is shaped by data availability and technical feasibility. By foregrounding professional ethos as an evaluative lens, this article offers a practice-centered approach to assess when AI may support, distort, or fail to address the aims of healthcare.

TAR DNA-binding protein 43 (TDP-43) inclusions are defining pathological features of frontotemporal lobar degeneration (FTLD) but are also often observed in Alzheimer's disease (AD) and primary age-related tauopathy (PART). TDP-43 in AD is either associated with cognitive impairment or a protective-life prolonging impact, and yet the localization, cellular and fragment characteristics of TDP-43 need to be determined. We investigated the relationships between TDP-43 volumetric inclusion burden in low likelihood AD (lAD) and definite PART by immunostaining against phosphorylated TDP-43 (pTDP-43), TDP-43 C terminal (TDP-C) and TDP-43 N-terminal (TDP-N) fragments combined with 3D confocal imaging taken from eight regions: amygdala (basolateral [amygdala-BL] and centromedial amygdala [amygdala-CM]), the hippocampus (Cornu Ammonis [CA]-1, CA2/3, CA4, dentate gyrus [DG] and subiculum [SUB]) and entorhinal cortex (ERC) and artificial intelligence (AI)-based segmentation via object recognition, reconstruction and quantification. We found amygdala-CM in lAD and PART to have the overall greatest burden of pTDP-43 whereas TDP-N burden in amygdala-BL of PART cases was greater than other TDP-43 fragments. There was no difference in TDP-43 burden in hippocampal subfields in PART. However, CA2/3 region showed greater pTDP-43 burden while TDP-N stood out in DG and SUB. Multiple comparisons among the groups revealed that TDP-C was the only fragment showing differences among PART and lAD in CA2/3, DG and SUB regions. Overall, unbiased AI-based volumetric burden analysis pipeline demonstrated unique fragment aggregation patterns in the neurodegenerative processes of PART and AD.

暂无摘要（点击查看详情）

暂无摘要（点击查看详情）

暂无摘要（点击查看详情）

暂无摘要（点击查看详情）

This review explores how e-health interventions are utilized within neonatal intensive care units to support nursing practice, with particular attention to their implementation areas, influence on care outcomes, and overall value for nursing services. Following PRISMA 2020 guidelines, 24 eligible studies were identified through a systematic search of four databases and appraised using design-specific quality assessment tools. The review synthesized findings from 24 studies investigating the integration of e-health technologies into neonatal nursing care. The evidence, spanning publications from 2000 to 2026, encompassed telehealth, mobile health applications, clinical decision support tools, digital learning strategies, and artificial intelligence-enabled systems. These interventions were mainly designed to enhance family participation, support care transitions following discharge, support breastfeeding, improve patient safety, and assist clinical decision-making processes. Overall, the findings suggest that e-health technologies contribute positively to several dimensions of neonatal care, particularly parental involvement, continuity of care, and breastfeeding support. Furthermore, emerging digital education and artificial intelligence applications appear promising for advancing nursing care delivery. Quality appraisal findings indicated a predominantly high-quality evidence base. E-health interventions can support family-centered care, continuity of care, patient safety, and clinical decision-making in neonatal intensive care units. Further high-quality multicenter studies are needed to evaluate their impact on clinical and nursing outcomes. The integration of e-health technologies into neonatal intensive care practice may enhance care delivery and support family engagement.

Retention of highly qualified physicians remains a critical priority in rheumatology and the broader medical field due to persistent physician shortages in the United States. Addressing physician well-being and burnout is therefore essential. An optimal approach to addressing burnout in the United States and internationally involves reimagining and redesigning care delivery. Physicians currently spend nearly equal time on administrative tasks and on direct patient care. A central challenge is to identify strategies that sustain professional engagement, enable providers to spend meaningful time with patients, and foster an environment that prioritizes provider well-being. Ambient artificial intelligence (AI) scribes are increasingly adopted for clinical documentation. However, evidence regarding their impact on productivity remains limited. These technologies require significant financial investment, and to offset associated costs, physicians are often expected to increase clinical volume by seeing additional patients. Although AI scribes may reduce documentation burden, their ultimate value depends on integration with broader initiatives to redesign clinical workflows, protect physician time, and restore meaning to patient care. Significant usability challenges persist. AI offers both substantial opportunities and notable risks. Thoughtful implementation within electronic health records and clinical workflows, combined with ongoing research on the impact of burnout, is required.

The increasing integration of artificial intelligence into learning environments has created new opportunities to examine how adaptive technologies influence fundamental psychological processes underlying language learning. This Study investigated whether learner-controlled voice interaction functions as a cognitive regulation mechanism that enhances second-language (L2) listening performance while reducing listening anxiety through greater learner agency and more efficient cognitive processing. A sequential explanatory mixed-methods design was employed with 57 Chinese university students learning English as a foreign language, including an experimental group (n&#xa0;=&#xa0;27) that used an intelligent voice assistant for adaptive listening support and a control group (n&#xa0;=&#xa0;30) that completed equivalent listening activities using conventional digital resources. Listening comprehension and foreign language listening anxiety were assessed before and after an eight-week intervention, followed by semi-structured interviews with selected participants. Analysis of covariance demonstrated that participants receiving adaptive voice-supported learning achieved significantly higher listening performance and significantly lower listening anxiety than those in the comparison group after controlling for baseline differences. Qualitative findings indicated that learner-controlled interaction facilitated cognitive regulation by reducing processing demands, strengthening perceived autonomy, and promoting self-regulated listening strategies. Rather than serving merely as a technological aid, adaptive voice interaction appeared to function as a psychological scaffold that supported both cognitive efficiency and emotional adaptation during demanding listening tasks. These findings extend contemporary perspectives on human-AI interaction by demonstrating how learner agency and cognitive regulation jointly contribute to successful language processing. The Study advances psychological research on technology-assisted learning by identifying adaptive voice interaction as a mechanism through which cognitive and affective processes can be simultaneously optimized in authentic learning contexts, while recognizing the methodological constraints associated with quasi-experimental classroom research.

Elucidating the internal representation of language in the brain has major implications for cognitive science, brain disorders, and artificial intelligence. A pillar of linguistic studies is the notion that words have defined functions, often referred to as parts of speech. Here, we recorded invasive neurophysiological responses from 1,801 electrodes in 20 patients with pharmacologically resistant epilepsy while they were presented with two-word phrases consisting of an adjective and a noun. We observed neural signals that distinguished between these two parts of speech. The representation of parts of speech showed invariance across visual and auditory presentation modalities; robustness to word properties such as length, order, frequency, and semantics; and even generalization across different languages. The selective signals were circumscribed within a small region in the left lateral orbitofrontal cortex. This selective, invariant, and localized representation of parts of speech extends classic fronto-temporal language models, providing a target for mechanistic and causal tests of how the brain represents the basic building blocks of language, and introduces a systematic approach to elucidate the orchestration of more complex aspects of language.

Identifying patients with glaucoma who are at risk of rapid disease progression is crucial to preventing vision loss. We aimed to develop and externally validate G-PROG, a deep learning model that predicts 2-5-year glaucoma progression from baseline colour fundus photographs (CFPs). G-PROG was trained and validated on data from a single centre (UZ Leuven, Leuven, Belgium); the other datasets (Brussels, Belgium; Li&#xe8;ge, Belgium; Tampere, Finland; Mainz, Germany; and Hangzhou, China) served as external test sets. Across six glaucoma departments, we analysed 161&#x2009;827 fundus images from 127&#x2009;962 visits (13&#x2009;913 patients), totalling 128&#x2009;021 eye-years of follow-up. Progression was defined by the G-RISK slope, calculated via within-eye linear regression on longitudinal G-RISK predictions over follow-up intervals of 2-5 years. G-RISK is a previously validated deep learning model that quantifies glaucomatous optic nerve damage from CFPs. We trained 20 G-PROG configurations with varying inclusion criteria applied to the number of visits, image quality, time between visits, and G-RISK at baseline. Performance was evaluated using the area under the receiver operating characteristic curve (AUC), the coefficient of determination (R2), and explained variance score (EVS). G-RISK slope as a progression biomarker was validated against the visual field mean deviation (MD) slope and average retinal nerve fibre layer thickness (RNFL) slope. Significant AUC values were obtained in 18 out of 20 model configurations, with internal validation reaching a maximum AUC of 0&#xb7;98 (95% CI 0&#xb7;97-1&#xb7;00) across follow-up intervals (2-5 years). In glaucomatous eyes with a baseline G-RISK exceeding 0&#xb7;6, the maximum AUC was 0&#xb7;92 (0&#xb7;85-0&#xb7;98). For external validation, the predictions from the eight top-performing configurations (selected based on positive R2 and minimal discrepancy between R2 and EVS in internal validation) were averaged. Maximum AUC values ranged from 0&#xb7;74 to 0&#xb7;86 across the five test datasets. G-RISK slope showed significant agreement with established progression markers, with maximum AUCs of 0&#xb7;82 for MD slope and 1&#xb7;00 for average RNFL slope. Externally validated across five international cohorts, G-PROG predicts 2-5-year glaucoma progression from baseline CFPs. Prospective evaluation is warranted to assess whether G-PROG can improve risk stratification and resource allocation in glaucoma care. This work was funded and supported by grants from the National Medical Research Council, National Research Foundation Singapore, National Health Innovation Centre Singapore, SingHealth and Duke-NUS, Duke-NUS, the Singapore Eye Research Institute and Nanyang Technological University and the Singapore Eye Research Institute, the Competitive Research Funding of the Pirkanmaa Wellbeing Services County, the LUX-Foundation for Glaucoma Research, state funding for university-level health research at Tampere University Hospital, Wellbeing Services County of Pirkanmaa, the Tampere University Hospital Support Foundation, and the Belgian Ophthalmology Cooperation in Clinical Sciences initiative hosted by the Funds for Research in Ophthalmology.

The increasing sophistication of cyber threats has led to the identification of some major shortcomings associated with honeypots, which include staticness, inflexibility, and vulnerability to fingerprinting. The proposed work aims at overcoming the aforementioned shortcomings by creating an Explainability-Driven Adaptive Cyber Deception Control System capable of engaging in intelligent, interactive interactions with cyber attackers. The key goal of the proposed solution is to improve threat intelligence gathering and deception efficiency by leveraging the benefits of adaptability and explainability. Machine learning, XAI, behavioral profiling, and environment mutation are the four key components that form the backbone of the proposed pipeline system. A Random Forest classifier is used for classification of normal and malicious sessions based on behavioral features at the level of commands. An explainability-driven metric known as the Feature Dominance Deception Index (FDDI) is developed to guide deception approaches, whereas Behavioral Convergence Score (BCS) is considered to assess behavioral convergence of attackers. Intent recognition using kill chain methodology allows generating responses in context-dependent fashion, while the mutation engine creates unique environments in each session to prevent fingerprinting attacks. Furthermore, Reinforcement Learning (RL) layer based on Q-learning is added to the framework to adaptively make decisions by learning the best possible deception tactics over multiple sessions. The Deception Quality Score (DQS) metric is used to measure the quality of deception within each session. Moreover, the UNSW-NB15 network intrusion data set is employed for validating the proposed model. Through benchmarking based on the generated behavioral dataset, the Random Forest-based behavioral profiler yielded a classification accuracy of 90.0%, recall of 85.7%, and an F1-score of 92.3%. Thereafter, the end-to-end deployment of the proposed framework through Cowrie honeypot sessions yielded better deception effectiveness, giving a framework-level attack classification accuracy of 90.0% and a 77.0% improvement in threat intelligence extraction per session than baseline Cowrie deployment. Kill chain stages were identified for the evaluated cases, deception goals were accomplished for all sessions under testing, fingerprinting efforts by the attacker were unsuccessful, and high-quality deception was maintained. The reward per session for the RL agent ranges from +&#x2009;0 to +&#x2009;14.0 for different session types, resulting in the formation of a converged Q-table containing values of 21 out of 90 possible states. Additionally, the technique ensures the resistance against honeypot fingerprinting, and demonstrates resistance against evaluated fingerprinting attempts. As far as it is currently known, few previous works can be found which have managed to include explainable scoring, convergence of behavior analysis, adaptive control, environment mutation, and reinforcement learning into one cyber deception framework. The presented framework manages to incorporate all of these features while still preserving transparency and adaptability during the whole process of deception. The research makes advances in the current state-of-the-art research by enabling passive honeypots to become intelligent autonomous systems for detecting cyber threats.

Patient safety investigation reports support organizational learning only when they are complete, usable, and sufficiently detailed. Conventional free-text reports are often inconsistent and may omit information needed for review and learning. Project NARRATE (Nursing AI-Refined for Accurate Transcription of Events) is a nursing-led ambient artificial intelligence workflow that uses prompts aligned with the World Health Organization Minimal Information Model for Patient Safety Incident Reporting and Learning Systems, Situation-Background-Assessment-Recommendation output, and visible missing-information cues to support structured supervisor reporting. This study aimed to compare the completeness and narrative quality of conventional and NARRATE-period supervisor investigation reports for falls and medication administration-related incidents. We conducted a retrospective pre-post document-quality study at a tertiary academic medical center in Singapore. We reviewed 150 deidentified completed supervisor investigation reports: 75 conventional reports from June to August 2025 and 75 confirmed NARRATE reports from January to March 2026. NARRATE use was voluntary, and recorded use represented approximately 40% of eligible postimplementation reports. Two blinded reviewers rated reports using a World Health Organization (WHO)-aligned completeness checklist and an adapted 8-domain Physician Documentation Quality Instrument (PDQI). Report-level comparisons were adjusted for repeated reports by the same supervisor using random-intercept linear mixed-effects models. A stratified 60-report plain-paragraph rerating examined whether visible structure influenced ratings. All 150 reports were analyzed. Unadjusted mean WHO total completeness was 11.81 (SD 3.39) for conventional reports and 13.61 (SD 2.54) for NARRATE reports; the unadjusted difference was 1.80 points, and the cluster-adjusted mean difference was 1.95 (95% CI 0.91-3.00; P<.001). The adapted PDQI mean was 3.61 (SD 0.52) and 4.13 (SD 0.34), respectively; the unadjusted difference was 0.52 points, and the cluster-adjusted mean difference was 0.53 (95% CI 0.37-0.69; P<.001). In the plain-paragraph sensitivity analysis, the completeness advantage remained (adjusted mean difference 1.70, 95% CI 0.27-3.14; P=.02), as did the adapted PDQI mean advantage (adjusted mean difference 0.25, 95% CI 0.06-0.43; P=.009). Explanation, organization, and comprehensibility remained significantly higher after deformatting; actions were borderline (P=.05), and synthesis, internal consistency, and fairness/balance were not statistically significant. Among voluntary early adopters, NARRATE use was associated with more complete reports and higher adapted PDQI mean scores after accounting for supervisor clustering. Because recorded use represented approximately 40% of eligible postimplementation reports and users self-selected, findings may reflect adopter and supervisor characteristics. Results support the structured workflow as a whole, not any single AI component, and do not demonstrate downstream patient-safety effects. Confirmatory evaluation under broader adoption with a concurrent, reliably classified comparison group is needed.

Administrative tasks substantially limit the time available for direct patient care in dentistry. The use of artificial intelligence (AI) can reduce administrative workload and improve practice efficiency, however its integration into routine dental practice remains limited. Existing research often focuses on specific AI applications or surveying digitally proficient practitioners, leaving practices with lower levels of digitalization underrepresented. This exploratory study establishes a baseline assessment of digitalization and AI readiness in Mecklenburg-Western Pomerania, Germany. German dental practices with varying degrees of digital maturity were included to assess existing digital infrastructure, current and intended AI use, and perceived barriers to adoption. A questionnaire-based survey was conducted among all dental practice owners in the German federal state of Mecklenburg-Western Pomerania. The exploratory survey assessed digital infrastructure, AI use, and perceived barriers. The data was analysed descriptively. A total of 163 dental practice owners participated, including a substantial proportion from rural regions (55%). Most practices reported exclusively digital patient records and appointment management, while the use of advanced digital tools varied. A minority had already implemented AI applications, which included support for administrative and diagnostic tasks, documentation, and radiographic evaluation. The most frequent reported barriers to AI adoption were perceived additional costs, concerns about increased workload, and limited familiarity with AI applications. AI adoption in dental practices in a low-density region remains limited, with practitioners favouring tools that reduce administrative workload. While digital infrastructure is increasingly available, barriers such as financial risk, implementation effort, and limited familiarity persist.

The integration of artificial intelligence (AI) into athletic training is accelerating, yet its psychological implications for athletes remain insufficiently understood. Drawing on the transactional model of stress and the stress-buffering framework of mindfulness, this study examined whether mindfulness training can mitigate adverse psychological responses associated with AI-assisted training. Using a randomized controlled factorial design, 160 collegiate athletes were assigned to AI-assisted training or standard training, with or without concurrent mindfulness intervention, and assessed at baseline, week 4, and week 8. Athletes exposed to AI-assisted training without psychological support exhibited increases in perceived stress and AI dependence over time. In contrast, these stress increases were substantially attenuated when mindfulness training was implemented alongside AI-assisted training. A significant AI &#xd7; Mindfulness &#xd7; Time interaction emerged for perceived stress at post-intervention, and difference-in-differences analyses corroborated a robust buffering effect. Mediation analyses further indicated that mindfulness training reduced stress partially through enhancing mindful awareness; a three-wave cross-lagged analysis showed that mindful awareness and stress were reciprocally related over time, with the hypothesized awareness-to-stress pathway remaining robust. Together, these findings suggest that AI-assisted training introduces a distinct form of evaluative pressure, and that mindfulness training may serve as an effective psychological buffer during the adoption of continuous algorithmic performance evaluation systems.

Accurate assessment of the critical view of safety (CVS) is essential for preventing bile duct injuries during laparoscopic cholecystectomy. Existing artificial intelligence approaches primarily rely on static frame-level analysis and often fail to capture the temporal evolution of surgical scenes, limiting their ability to provide reliable and context-aware safety assessment. To address this challenge, we propose TempoSafe-CVS, a temporal multi-scale framework for automated CVS assessment in surgical videos. The proposed architecture integrates complementary visual representations through a Swin Transformer-based global context encoder, a ResNet-based local feature extractor, and a structure-aware convolutional module. These multi-scale features are combined and processed using temporal sequence modelling and spatio-temporal reasoning to capture both visual and temporal dependencies across surgical sequences. Furthermore, a unified multi-task prediction framework jointly estimates CVS safety status, procedural progression, anatomical structure visibility, and clinically relevant C1/C2/C3 criteria. Experiments conducted on the Endoscapes benchmark dataset demonstrate the effectiveness of the proposed approach, achieving 79.6% AUC-ROC for safety assessment, 81.5% average balanced accuracy for C1/C2/C3 criteria classification, and a mean absolute error of 0.187 for progression estimation. Comparative evaluations show consistent improvements over existing CVS assessment methods, highlighting the benefits of temporal reasoning and multi-scale visual representation learning. Qualitative analyses further demonstrate the interpretability of the framework through temporally consistent and anatomically grounded predictions. The proposed framework advances intelligent surgical video understanding by combining temporal sequence reasoning with multi-scale visual analysis, offering a potential solution for explainable and context-aware decision support in safety-critical surgical environments.