Stroke, a leading cause of mortality, manifests as ischemic (87%) or hemorrhagic (13%), demanding rapid intervention to mitigate irreversible damage. Despite advances in artificial intelligence, systematic reviews addressing the integration of large language models and natural language processing into clinical stroke care remain limited. Such a review is critical given large language models' potential to overcome traditional natural language processing limitations, thereby enhancing risk prediction and decision support. We proposed a systematic review aimed to comprehensively analyze applications in stroke care. After searching 6 databases, 2991 records were screened, yielding 65 eligible studies. Results were structured quantitatively (impact factor trends, publication distribution) and qualitatively across 4 dimensions: Study Purposes (eg, risk modeling, decision support), Data Sets/Key Findings, Limitations, and Future Directions. Large language models demonstrated strong capabilities in automated data extraction from clinical notes (accuracies of 93.5%-95.1%) and report summarization. However, majority of studies (94%) lacked external validation. Most were limited by single-center, retrospective designs (62%) and used private data sets (85%), raising concerns about generalizability. Common failure modes included model hallucinations, performance degradation on external data, and infrastructural barriers to clinical integration. Future efforts must prioritize multicenter, prospective validation (82% of studies) to ensure model robustness and generalizability across diverse populations. Pathways for clinical translation include developing interpretability techniques to build clinician trust. Technical refinements for hallucination mitigation (98% of studies) and real-time integration of multimodal data are necessary to enhance predictive power. Addressing data heterogeneity and ethical concerns remains as gaps. This review highlights the potential of large language models in stroke care, encompassing tasks from risk prediction to workflow automation. Realizing this potential requires a shift from proof-of-concept studies to rigorously validated, clinically integrated systems. The field demands scalable, equitable, and transparent artificial intelligence solutions that are codeveloped with clinicians. These are needed to overcome existing methodological and translational barriers.

Lewy body disease (LBD) and Alzheimer's disease (AD) are the most common causes of cognitive decline and dementia and are associated with characteristic alterations in resting-state electroencephalographic (rsEEG) activity. This multicenter exploratory study investigated periodic and aperiodic rsEEG features in patients with cognitive decline due to Lewy body disease (LBCD) and Alzheimer's disease (ADCD), compared with cognitively unimpaired older adults (Nold), and examined the clinical relevance of these markers in LBCD. A total of 140 LBCD, 135 ADCD, and 118 Nold datasets from the PDWAVES archive underwent spectral parameterization to decompose rsEEG power spectra (1-30 Hz) into periodic peaks and aperiodic background activity. Both clinical groups showed a significant slowing of the individual alpha frequency (IAF), more pronounced in LBCD, along with reduced periodic alpha and beta power reflected in a lower vigilance index. The aperiodic exponent was elevated in both groups, and the aperiodic offset was also higher in LBCD, suggesting steeper spectral profiles consistent with increased inhibitory cortical tone. Within the LBCD group, poorer cognition was associated with higher low-frequency alpha power, whereas better cognition was predicted by higher high-frequency alpha power. A reduced vigilance index was associated with the presence of visual hallucinations, while no associations emerged for other symptoms. These findings suggest that combined periodic and aperiodic rsEEG features may provide relevant markers of altered vigilance regulation in LBCD. Future studies should evaluate whether these EEG markers can inform targeted interventions, such as neuromodulatory or audiovisual stimulation, to stabilize quiet-vigilance states and improve clinical outcomes.

Healthcare systems in the United States are now mandated to provide patients with immediate access to medical results. Access to complex results prior to appropriate clinical follow-up may cause confusion and fear, especially in populations with limited health literacy and English language proficiency. In this study, we demonstrate a use case of large language models (LLMs) to simplify and translate complex medical information for patients. We collected a consecutive series of brain MRI (N = 200), head CT (N = 100), and spine MRI (N = 200) reports, prompting a GPT-4o model to simplify these reports to a middle-school reading level and translate a subset of them to Spanish. Critically, two independent neuroradiology raters compared the simplified report impressions to the original report impressions, and found low rates of hallucinations, omissions, and imprecisions. The reading grade of the report impressions decreased from approximately college to approximately middle-school level, and readability improved across several text complexity metrics. Additionally, simplified impressions were classified as more neutral and less fearful. Machine generated translations of simplified impressions were statistically indistinguishable from manual translations as assessed by certified interpreters. In sum, we present a scalable, patient-friendly approach to promote healthcare literacy and engagement.

Hallucination risk and trustworthiness of a generative artificial intelligence system play an important and vital role in daily-life scenarios. The terms Hallucination risk and trustworthiness are very famous for the generative artificial intelligence techniques to confirm that the final results or outputs are safe, accurate, and reliable for users or customers, particularly in education, finance, and healthcare. The main purpose of this manuscript is to evaluate the best and worst decisions among the decisions of Gemini (Google DeepMind), Copilot (GitHub/Microsoft), LLaMA (Meta), Claude (Anthropic), and ChatGPT (OpenAI) based on the following criteria, such as data privacy, explainability, Hallucination risk, reliability, and bias control. Therefore, the system of interval-valued m-polar fuzzy-multi-attribute border approximation area comparison technique for algebraic information is developed. The invented model is very famous and reliable for coping with awkward and complex information. Therefore, for the valuation or the construction of the derived technique, this study aims to develop the novel technique of interval-valued m-polar fuzzy information and also investigate their operations based on algebraic norms for the construction of the averaging operator and geometric operator. Lastly, this study also works on the analysis of the numerical examples and the implementation of the comparative analysis among proposed and prevailing frameworks to derive the validity and effectiveness of the introduced approaches.

Bioinformatics is entering a new phase characterized by the integration of universal biological models and multi-agent systems to enable end-to-end scientific discoveries. This review argues that the next paradigm shift will go beyond traditional predictive models and generative artificial intelligence (AI) toward agentic AI: systems capable of planning, acting through tools, reflecting on results, and iterating until a goal is achieved. We first examine recent foundational models that produce transferable representations across omic modalities, such as scGPT, Nicheformer, and EpiAgent, and discuss their architectural choices, training regimes, and interpretability constraints. We then analyze biomedical agent frameworks through their main components (planning, action, reflection, and memory), highlighting representative systems such as ClinicalAgent and Biomni that operationalize these ideas in controlled environments. Next, we focus on hypothesis validation mechanisms, including retrieval-augmented generation for evidence grounding, sequential statistical testing, and benchmarking methodologies designed to quantify robustness and reproducibility. Finally, we summarize emerging applications in drug discovery and personalized medicine, from molecular literature analysis and protocol automation to drug repurposing for rare diseases and closed-loop synthesis. We conclude by outlining the main challenges ahead, namely hallucinations, interpretability, systemic biases, integration with clinical infrastructures, and regulatory and ethical requirements, and propose a roadmap for the development of scientific agents that are not only high-performing but also reliable, verifiable, and implementable in real biomedical contexts.

Large language models (LLMs), built on transformer architecture, have emerged as a fundamental tool in natural language processing and contextual reasoning, and have been extended to multimodal data interpretation, which has been termed large multimodal models (LMMs). Radiology as a medical discipline, having undergone full digital transformation over the past two decades, is uniquely positioned at the forefront of medicine to benefit from this recent technological advancement. The integration of LLMs and LMMs into radiology workflows has demonstrated promise in improving reporting efficiency, decision support, and clinical communication. However, real-world adoption remains limited due to concerns about reliability, hallucinations, drift, workflow disruption, medicolegal uncertainty, and the absence of standardized integration to existing clinical systems/infrastructures. This review highlights the need for transparent model behavior, standardized software integration tools, and high-quality radiologist-curated local datasets. While early studies suggest that LLMs and LMMs may reduce cognitive load and enhance reporting efficiency, their clinical value depends on alignment with radiologists' needs, well-planned deployment, and rigorous evaluation and maintenance. We conclude with suggestions toward effective integration of LLMs and LMMs into modern, constantly evolving radiology practices.