Reliability of AlphaFold2 predictions is mainly assessed using the predicted Local Distance Difference Test (pLDDT). For model organisms, 30-40% of residues fall into the low-confidence pLDDT range. Moreover, pLDDT sometimes fails to flag physically implausible structures. This raises two questions: can more robust reliability indicators be identified, and do unreliable predictions share common structural or biophysical features?

Artificial intelligence (AI) is rapidly being adopted in education in the health care professions, including in palliative care. Yet existing AI primers for health professions education (HPE) are not specific to palliative care (PC) and overlook the relational, prognostic, and cultural sensitivities central to the field. This narrative primer addresses that gap. Informed by a review of the literature, it equips PC educators with practical guidance for responsibly harnessing AI. We first introduce foundational AI concepts relevant to educators and clinicians, including machine learning (ML), large language models (LLMs), generative AI (GenAI) and agentic AI. We then trace a progression from general HPE use, such as study support, assessment, and AI-enhanced simulation, to PC-specific applications in curriculum design, serious-illness communication training, and interprofessional teamwork. Throughout, we situate the risks where they arise, with attention to concerns most consequential for PC: bias, communication integrity and hallucination, data privacy, and over-reliance on AI, in a field where relational, humanistic practice and nuanced communication are central. Guiding principles of ethics, equity, and patient-centeredness anchor the discussion. We close with concrete implications for educators and curriculum development: building AI literacy, establishing governance and appropriate-use policies, and verifying AI-generated outputs against trusted sources. The aim is an educator-AI partnership that safeguards what is essential in PC: compassionate, dignified, patient-centered decision-making and care.

Quantitative systems pharmacology (QSP) models require calibration data from literature, yet manual curation is inconsistently documented and large language model (LLM) extraction can hallucinate values and fabricate citations. We present MAPLE (Model-Aware Parameterization from Literature Evidence), which uses structured validation schemas as a collaboration interface between LLMs and modelers. Two schemas span two scales: the SubmodelTarget schema for isolated experiments constraining individual parameters, and the CalibrationTarget schema for clinical and in vivo endpoints constraining the full model. Both separate data extraction from modeling decisions, recording every value with full provenance. Targeted validators catch characteristic LLM errors by matching values to source snippets, resolving DOIs, and executing code. For a pancreatic ductal adenocarcinoma QSP model, we used MAPLE to extract and curate 37 SubmodelTargets and 45 CalibrationTargets. Before any human review, the validators triggered 50 automated retries; every value carries a direct quote from its source and a verified citation; and 11 of 19 parameters are supported by more than one independent source. The LLM drafted usable forward models and code from context, while the modeler supplied the context and scientific judgment it cannot infer, revising forward-model choices in 65% of SubmodelTargets, priors in 46%, and source relevance in all files. This evaluation covers one model in one disease area, by a single group, so it characterizes the framework rather than establishing how broadly it generalizes. MAPLE records the modeler's reasoning in a form that can be re-run and independently checked, so it is not lost when the modeling team changes.

As generative artificial intelligence (GenAI) becomes deeply embedded in cognitive workflows, its inherent probabilistic generation mechanisms and the possibility of machine hallucinations may complicate the stability of semi-technical users' trust. This study examined how semi-technical users' interaction trust toward generative AI agents is associated with perceived explainability, perceived intention alignment, and perceived sense of agency, while also considering the moderating roles of perceived task complexity and domain self-efficacy.

Obsessive-compulsive disorder (OCD) may be present with distressing intrusive thoughts, urges, or images. When the content is violent, sexual, religious, or linked to magical thinking, severe OCD can be mistaken for psychosis. We present a 19-year-old man with a six-year history of ego-dystonic intrusive harm-related thoughts and images focused on family members, along with checking, touching, counting, reassurance seeking, religious mental rituals, and sexual obsessions. Symptoms began at age 13 after he read about asthma, and later developed a recurrent fear that his maternal grandmother might develop the illness. He recognized these thoughts and images arising from his own mind, experienced them as distressing and excessive, and used rituals to reduce anxiety. Examination did not elicit hallucinations, externally located perceptions, thought insertion, withdrawal, broadcasting, passivity phenomena, behavioral disorganization, or mood, substance-related, or neurological syndrome. The diagnosis was obsessive-compulsive disorder, International Statistical Classification of Diseases and Related Health Problems10th Revision (ICD-10), F42.2 mixed obsessional thoughts and acts. Fluvoxamine 50-150 mg/day was associated with a subjective 10%-20% improvement without documented adverse effects; the treatment was later switched to fluoxetine because of cost-related factors. Long-term follow-up and standardized severity scores were unavailable. This case highlights the importance of assessing thought ownership, resistance, insight, conviction, and the anxiety-reducing function of rituals before labeling vivid intrusive mental images as hallucinations or delusions.

BackgroundNeuropsychiatric symptoms (NPS), including agitation, depression, anxiety, psychosis, apathy, and sleep disturbances, are highly prevalent in dementia and significantly contribute to carer burden, accelerated cognitive decline, and increased healthcare utilisation. Despite their clinical relevance, NPS often remain under-recognised and sub-optimally managed. This review synthesises current evidence on the prevalence of NPS in dementia, their functional and psychosocial impacts, and pharmacological and non-pharmacological management approaches.MethodsA systematic search of PubMed, Embase, PsycINFO, and the Cochrane Library (2013-2025) identified observational and interventional studies, which were systematically reviewed and synthesised. Data were independently extracted and quality-assessed.ResultsEighty-two studies (n = 67,945) were included. Neuropsychiatric symptoms were highly prevalent, affecting approximately 78% of individuals with dementia. The most common symptoms were apathy (39%), agitation (35%), and depression (32%), with variation across dementia subtypes. Neuropsychiatric symptoms were associated with functional decline, increased risk of institutionalisation (aHR = 2.15), higher caregiver burden, and increased healthcare costs (approximately 38%). Multicomponent non-pharmacological interventions reduced symptom severity by 18-28%, while pharmacological treatments showed modest efficacy and were associated with increased adverse events. The certainty of evidence ranged from high (prevalence estimates) to moderate (psychosis and hallucinations) and low to moderate (pharmacological interventions).ConclusionNeuropsychiatric symptoms are highly prevalent in dementia and have a substantial impact on patients, carers, and healthcare systems. Individualised non-pharmacological interventions should be first-line, with pharmacological treatments reserved for severe or refractory cases. Early screening and intervention are essential.

Large language models (LLMs) are increasingly explored in radiology, yet concerns persist regarding hallucination and lack of factual grounding. Retrieval-augmented generation (RAG) seeks to address these limitations by coupling generative models with external knowledge retrieval. We conducted a scoping review to characterize how RAG systems have been applied in radiology and medical imaging. A systematic search of PubMed, Embase, Scopus, IEEE Xplore, and arXiv identified 45 studies implementing RAG-based approaches in radiology-related tasks. In terms of clinical tasks, RAG was most commonly applied to radiology report generation and question answering. Dense retrieval strategies predominated, while sparse, hybrid and proprietary retrieval approaches were less frequent. External knowledge sources most frequently comprised biomedical literature databases and clinical guidelines. Applications were heavily skewed toward chest radiography and X-ray-based tasks, with relatively few studies addressing CT, MRI, PET, ultrasound, or under-represented subspecialties such as pediatric radiology and neuroradiology. Most comparative studies reported task-specific performance gains with RAG over non-retrieval baselines, and a small subset reported performance comparable to trained radiologists or state-of-the-art models. However, hallucinations and errors persisted, and heterogeneity across studies limited the generalizability of these findings. Evaluation practices largely relied on automated accuracy or text-overlap metrics, with limited use of standardized expert evaluation and minimal assessment of safety, bias, computational efficiency, or clinical utility. Overall, while RAG shows promise for improving factual grounding in radiology AI, current evaluation paradigms likely overestimate real-world clinical readiness. Future work should prioritize retrieval quality, clinically grounded evaluation, safety-critical error analysis, bias assessment, and deployment-relevant efficiency metrics to enable responsible clinical translation.

Artificial intelligence (AI) is increasingly evolving from a research technology into a tool for everyday clinical practice. While early applications primarily focused on medical image analysis, generative AI systems and large language models are now available for a wide range of clinical and administrative tasks. These include medical documentation, literature review, guideline-based knowledge management, patient communication, and workflow optimization. At the same time, diagnostic and therapeutic applications continue to evolve. AI-assisted systems support radiological and pathological image interpretation, risk stratification, and clinical decision-making processes. Despite considerable opportunities, important limitations remain. AI hallucinations, algorithmic bias, data protection requirements, and regulatory considerations necessitate continuous human oversight and critical evaluation. Therefore, the long-term success of AI will depend not only on technological performance but also on its responsible integration into existing healthcare structures. This review provides a practice-oriented overview of current and future AI applications in urology and discusses opportunities, limitations, and prerequisites for safe implementation in clinical practice and hospital care.