Multi-model large-scale AI framework for avian influenza surveillance and preparedness: Harnessing large language models to enhance risk communication, real-time decision support, and public health response strategies.

Generally, gastroenterology and digestive endoscopy units commonly face constraints. It may be due to a lack of equipment, poor scheduling and inadequate inventory management. These issues not only affect patient throughput but also increase the workload for medical staff. Thus considering the current advancements in artificial intelligence we can work towards new opportunities to enhance clinical pathway and resource optimization to overcome the problems. The purpose of this systematic review is (i) to study the work done by the researchers in the field of gastroenterology and endoscopy services and LLM-powered chatbots and (ii) to check the possibilities of combining LLM-powered chatbots with traditional and dynamic healthcare inventory models. Common platforms like Google Scholar, PubMed, Scopus, CrossRef and Web of Science were explored for relevant recent literature. The search was conducted using keywords "endoscopy," "chatbots," "large language models," "inventory models," and "gastroenterology." 222 records were collected in total, among which 91 met the required criteria and were subjected to a transparent review process following PRISMA guidelines. Technical, clinical and organizational issues of LLM implementation in healthcare were addressed in the eligible studies. Slightly less than half of the received papers described predictive scheduling and workflow optimization systems, others addressed research on data management, documentation or interaction with patients. Few of these discussed about moral or legal considerations. The analysis focuses of ability of LLM chatbots to support product forecasting, equipment utilisation, waste reduction and real-time decision support in endoscopy units when integrated with predictive inventory models. The amalgamation of LLM chatbots with inventory management is an effective way to achieve intelligent and sustainable activities in gastroenterology. Even though some problems will arise with this amalgamation, especially when it comes to data interoperability, ethics and validation. These two fields can be combined to work towards more effective, flexible and patient-centric environment.

Hybrid deep learning architecture for skin disease classification

Heart failure with preserved ejection fraction (HFpEF) is a prevalent and heterogeneous syndrome with limited therapeutic options, making accurate risk stratification essential yet challenging. Traditional tools such as the H2FPEF and HFA-PEFF scores incorporate few variables and demonstrate modest prognostic performance. Machine learning (ML) offers enhanced risk prediction by integrating multidimensional clinical, imaging, biomarker, and molecular data. This review summarizes current ML applications in HFpEF, including random forests, gradient boosting, support vector machines, and deep learning, highlighting their superior discrimination and ability to reveal phenotypic subgroups with distinct outcomes. We also address practical considerations such as interpretability, real-world validation, and integration into clinical workflows, as well as challenges related to data bias, generalizability, and regulatory requirements. Future opportunities include real-time clinical decision support, digital health integration, and interventional ML to guide personalized therapy. ML holds significant potential to advance precision care and improve outcomes in HFpEF.

Advancing continuous in-situ quantification of microbial contamination in environmental waters using tryptophan-like fluorescence-Sensor design and validation.

This is Part II of a two-part series examining artificial intelligence (AI) in colorectal surgery. Part I established foundational concepts and clinical applications. Implementation, however, requires understanding research methodologies, available resources, and the specific challenges currently limiting widespread adoption. This will be the focus of Part II. To examine artificial intelligence's transformation of surgical research, provide practical implementation resources, address adoption challenges, and explore future directions in colorectal surgery. Comprehensive literature review focusing on artificial intelligence research methodology, implementation barriers, educational resources, and emerging technologies relevant to colorectal surgeons. Artificial intelligence streamlines clinical trial design through predictive modeling and natural language processing, reducing enrollment challenges that contribute to failed or inadequate trial accrual. Machine learning enables heterogeneity analysis within clinical trials, identifying treatment-responsive subgroups. Foundation models unlock analysis of unstructured electronic health record data at scale. Professional societies and universities offer specialized artificial intelligence education programs, with open-access datasets facilitating research participation. However, implementation faces multifaceted challenges: technical infrastructure demands, with real-time processing requiring dedicated graphics processing unit clusters; regulatory frameworks struggling with continuously evolving algorithms; undefined liability distribution for artificial intelligence-assisted decisions; algorithmic bias risking healthcare disparities; and the "black box" problem limiting clinical trust. Economic barriers include substantial initial costs without clear reimbursement pathways. Future directions include multimodal artificial intelligence integrating imaging, genomics, and histopathology; cognitive robotic systems with real-time decision support; digital twin technology for patient-specific surgical simulation; and global surgical artificial intelligence networks enabling distributed learning across institutions. While artificial intelligence offers transformative potential for colorectal surgery research and practice, successful implementation requires addressing technical, regulatory, ethical, and economic challenges. The surgeon's evolving role demands both traditional expertise and computational fluency. Future advances in multimodal integration, autonomous systems, and global collaboration will fundamentally reshape surgical practice, but require thoughtful implementation prioritizing patient benefit and clinical value.

Total Cholesterol (TC) monitoring in blood serum is critically important for assessing cardiovascular risk and metabolic health. In this work, we report a nano-bioscience-enabled cholesterol-sensing portable platform that integrates copper oxide (CuO) nanoparticles and a hybrid microfluidic chip, along with smartphone-based chemiluminescence (CL) imaging and computational analysis. The hybrid microfluidic miniaturized chip, fabricated using PDMS material and Whatman filter paper, enables efficient reagent transport, on chip micromixing, and localized signal generation. CuO nanoparticles act as catalytic enhancers for the luminol-H₂O₂ chemiluminescence reaction initiated by cholesterol oxidase, resulting in strengthened and stable photon emission. Chemiluminescence signals were captured using a smartphone camera placed in a 3D-printed dark box and analyzed through a customized mobile application, Intensity Tracker application, for intensity evaluation proportional to cholesterol concentration. The proposed work gives a linear detection range for the cholesterol over a concentration of 0.06-1.5 mM, with a limit of detection of 0.05 mM and a limit of quantification of 0.191 mM, demonstrating strong analytical performance R² ≈ 0.98. Validation using human serum samples yielded recovery values ranging from 92 to 101%, confirming high accuracy and minimal matrix interference when compared with results from a standard biochemistry analyzer. The developed hybrid microfluidic chemiluminescence analyzer features a compact design, low cost, and user-friendly operation, presenting a promising solution for rapid, accurate, and decentralized cholesterol monitoring in point-of-care and resource-limited settings. Future work will focus on multiplexed biomarker detection, and integration with AI-based analysis for real-time clinical decision support.

Current advances and applications of artificial intelligence for smart municipal solid waste management: A review.

Sepsis is a potentially fatal systemic response to infection, in which early clinical intervention is critical to reduce mortality. This study presents a hybrid deep learning model that combines temporal and structural information from clinical data to improve early sepsis prediction. We used data from the PhysioNet/Computing in Cardiology Challenge 2019 to predict sepsis onset up to 12 hours in advance. We developed a hybrid model integrating long short-term memory (LSTM) networks and graph attention networks (GAT) to capture temporal dynamics and intervariable relationships. Performance was compared with three baseline models. To ensure robustness, all models were trained using five repeated train-test splits with different random seeds. The dataset includes 40 336 adult ICU patients. Of all the patients, 2932 developed sepsis during their stay. Each patient's data includes hourly data on 40 clinical variables, including vital signs, laboratory results and demographic information. The LSTM-GAT model achieved an area under the receiver operating characteristic curve (AUROC) of 0.853±0.005, an F1-score of 0.627±0.006 and a specificity of 0.872±0.007, outperforming baseline models. Despite being trained on fixed temporal windows, the model generalised well across multiple prediction horizons without retraining. By integrating temporal and structural representations, the proposed approach achieves improved predictive performance compared with the baseline. This capability may support earlier identification of high-risk patients and enhance timely clinical decision-making in critical care environments. The proposed model demonstrates the advantage of combining sequence and graph-based methods. It offers a promising tool for real-time clinical decision support in sepsis detection.

Early detection of malignancy is imperative, yet existing diagnostic approaches struggle to identify small peripheral lesions. This study evaluated a novel imaging modality, heat diffusion analysis, to assess its ability to differentiate between malignant and normal lung tissue. Considering that lung cancer is the leading cause of cancer-related mortality worldwide, lung tumors were induced in mice in a preclinical ex vivo model to evaluate the proposed technology. The HTOScan System was used to analyze the thermal characteristics of 60 sites from excised lungs, including normal and abnormal regions. The algorithm classified pixels as high- or low-risk for malignancy. The HTOScan System demonstrated a high accuracy of 97%, with 94% sensitivity and 98% specificity compared to the gold standard of histopathology. The technology successfully differentiated abnormal from normal tissue ex vivo based on differences in thermal diffusivity. This proof-of-concept study suggests that combining heat diffusion imaging techniques with machine learning algorithms could enable the HTOScan System to identify malignant lesions accurately with high confidence. The technique shows promise as a real-time decision support tool for cancer detection, pending further in vivo validation. This novel functional-imaging approach could improve the identification of peripheral lesions and the guidance of biopsies during bronchoscopy.

TRIAG: Tri-reinforced infused generative agents for financial risk compliance

This study presents and evaluates a real-time decision support system (DSS) for site-specific irrigation scheduling based on soil moisture forecasting with SWIM 2 (Sensor Wielded Inverse Modeling of a Soil Water Irrigation Model). The SWIM 2 framework integrates a soil water balance model with in situ sensor data and soil moisture samples through Bayesian inverse modeling to generate probabilistic 10-day soil moisture forecasts. We assess the performance of the soil moisture forecasts and the irrigation DSS by integrating the model parameter ensemble with either deterministic or ensemble-based probabilistic weather forecasts, providing insights into their benefits and trade-offs in real-time irrigation management. Both approaches resulted in high detection rate and accuracy in predicting water stress triggering the irrigation threshold. The full ensemble yielded slightly better reliability at longer lead times whereas the probability distribution of the soil moisture predictions at short lead times was dominated by the SWIM 2 parameter uncertainty. Simulation of different irrigation treatments using the calibrated SWIM 2 -based model illustrated and confirmed its potential for evaluating water use efficiency and crop response. Overall, this work illustrates the application and practical advantages of a probabilistic, ensemble-based modeling framework in supporting site-specific, data-informed irrigation strategies. • SWIM 2 enables real-time, site-specific irrigation advice using probabilistic soil moisture forecasts. • Soil moisture uncertainty shifts from model parameter to weather uncertainty dominance with longer lead times. • Deterministic and ensemble weather forecasts were compared in water stress predictions for irrigation support. • Simulated irrigation strategies reveal impacts on water use efficiency, irrigation efficiency, and productivity.

SafeDriveEdge: multimodal vision-language reasoning for real-time decision support in intelligent vehicles

IntroductionExtracorporeal membrane oxygenation (ECMO) provides life support for patients with refractory cardiac or respiratory failure. The complexity of ECMO management and associated mortality necessitates high-accuracy clinical decision-making systems. Artificial intelligence (AI) has emerged as a potential approach to address challenges in ECMO management, from patient selection to real-time assessment and outcome prediction.ObjectiveTo synthesize the current evidence of AI application in adult ECMO, addressing predictive modelling for patient outcomes, real-time decision support systems, and complication prevention, as well as the evolving regulatory challenges governing medical AI deployment in critical care settings.MethodsA narrative literature review was conducted across PubMed/MEDLINE, Embase, Cochrane Library, IEEE Xplore, and preprint servers (arXiv/medRxiv). The search strategy combined ECMO-relevant terms ("V-A ECMO", "V-V ECMO") with AI terminologies ("artificial intelligence", "machine learning", "deep learning", "digital twin"). Studies were included if they focused on adult cohorts (age ≥18 years) and were published in English between 2018 and 2025.ResultsThe review found several AI algorithms under development for different stages of ECMO therapy. AI algorithms have been developed to assist in the initiation, prognostication, complication detection, real-time control, and weaning of ECMO. However, none have been clinically translated thus far.ConclusionWhile AI for precision ECMO management is promising, several prerequisites remain unmet, including the integration of high-frequency device data, prospective external multicenter validation, and the development of robust regulatory frameworks. Securing these advances will bridge the gap between algorithm development and the clinical arena.

Prompt-to-policy: Leveraging large language models to guide deep reinforcement learning in public health emergencies.

Various high-tech technologies are used extensively around the world in sports competitions especially in athletic events where scientific real-time decision-making is essential to improve competitive efficiency and match outcomes. This research proposes the introduction of the concept of entropy by applying the ID3 algorithm with the use of the attribute entropy value change as the selection criterion to develop the decision tree model for real-time sports competition data processing. Meanwhile, an enhanced Monte Carlo tree search algorithm can select the maximum UCT function node to ensure the optimal solution. The study shows that there is a percentage of players who have used advanced strategies in their basketball games ranging from 0 to 0.04. The developed decision system can make real-time strategy evaluation for sports competitions, taking a decision-making time of about 3.13 seconds on average. In addition, the system makes a 74% win rate and 84% decision rationality which indicates that the decision system is quite ideal and can serve as a good reference in future sports competition real-time decision-making practices.

The exponential growth of unstructured biomedical data, comprising nearly 80% of healthcare information, has created both unprecedented opportunities and formidable challenges for extracting actionable insights from clinical notes, electronic health records, biomedical literature, and patient-reported outcomes. This review examines the evolution of AI-driven text mining from experimental application to clinical necessity, with emphasis on disease prediction, drug discovery, and clinical research. Literature encompassing natural language processing (NLP), transformer-based models, clinical case studies, and regulatory frameworks was systematically analyzed across oncology, cardiology, neurology, and pharmacovigilance domains. AI-enabled text mining demonstrates robust performance across multiple applications: disease prediction models achieve 67–98% accuracy in early diagnosis and risk stratification; transformer-based methods yield 80.6% F1-scores for drug–target interaction extraction; and adverse drug reaction detection from social media achieves 84.2% sensitivity and 98% specificity. In clinical research, systematic review timelines are reduced by up to 70%, and clinical trial recruitment screening requirements decline by nearly 80%. Real-time clinical decision support powered by large language models reduces diagnostic time from over 30 minutes to less than one minute, maintaining accuracy comparable to expert teams. Despite remarkable progress, challenges persist, including data heterogeneity, annotation quality, computational demands, translational gaps, algorithmic bias, and privacy concerns. Future directions include multimodal integration with genomics, imaging, and biosensors, explainable AI frameworks, and federated learning for collaborative research. AI-enabled text mining represents a transformative paradigm shift toward predictive, preventive, and personalized medicine, bridging the gap between exponential data growth and human cognitive limitations while improving patient outcomes and accelerating scientific discovery.

Global logistics enterprises generate vast transactional records that remain underutilised for strategic profitability management. This paper applies systematic Exploratory Data Analysis (EDA) and multi-dimensional feature engineering to a real-world operational dataset of 180,519 order records from APL Logistics (KWE Group), spanning 20,652 unique customers, 118 products across 50 categories, five global markets, and 23 order regions. The dataset encompasses total portfolio revenue of $36,784,734.31 and net profit of $3,966,902.97, yielding an overall profit margin of 10.78%. Critical findings include: 18.71% of orders (33,784 transactions) are loss-making; total discount expenditure ($3,730,378.40) constitutes 94.0% of net profit; the top 10% of customers (2,066 accounts) account for 49.1% of cumulative profit; a Pearson correlation of r = -0.0027 (p = 0.253) between discount rate and profit ratio — non-significant at the individual order level — coexists with a monotonic 25.2% aggregate-level margin decline from zero-discount to 21–25% discount bands; and a portfolio-wide late delivery rate of 54.83% imposes significant service-quality risk. An interactive six-module Streamlit dashboard is deployed to operationalise analytical outputs for real-time commercial decision support. The study advances the evidence base for profit-centric supply chain analytics and provides actionable recommendations for discount policy reform, customer value tiering, and shipping mode rationalisation

The rapid increase in hypertension and diabetes cases has created a strong demand for advanced healthcare systems that support early diagnosis and effective clinical decision-making. Managing these chronic conditions requires continuous evaluation of multiple patient attributes, which becomes difficult when performed manually. Traditional approaches rely heavily on basic statistical techniques and human interpretation, making them inefficient for handling large-scale data, identifying complex patterns, and providing real-time predictions. As a result, such systems often lack accuracy, scalability, and reliability, particularly in distributed healthcare environments. A key challenge lies in their inability to handle imbalanced datasets, perform multi-condition prediction, and enable seamless remote communication between systems. To address these limitations, this work proposes a real-time decision support system powered by Artificial Intelligence (AI) using a dual client–server architecture. The server handles data preprocessing, model training, and prediction using Machine Learning (ML) algorithms such as Complement Naive Bayes (CNB), Multinomial Naive Bayes (MNB), Perceptron, and a Tao Tree Classifier (TTC). Preprocessing methods include Label Encoding and K-Means Synthetic Minority Oversampling Technique (KMeans-SMOTE) to manage categorical data and class imbalance. A Flask-based Application Programming Interface (API) using Hypertext Transfer Protocol (HTTP) enables efficient communication between the client and server. The client system allows users to upload datasets, which are processed remotely to predict blood pressure categories and diabetes status. Lightning Memory-Mapped Database (LMDB) is used for secure and efficient data management. The proposed system ensures accurate multi-target prediction, real-time accessibility, and seamless device communication, ultimately improving healthcare services, reducing manual effort, and supporting better clinical decisions.

ABSTRACT Co-axial Wire Laser–Directed Energy Deposition (DED-LB/w) is an emerging metal additive manufacturing (AM) technique that combines near-zero material waste with high deposition efficiency. However, the co-axial wire-feeding configuration is prone to wire-tip dripping, which destabilises the melt pool and degrades build quality. Without timely in-situ intervention, dripping can lead to costly repairs, build interruptions and even tool damage, underscoring the critical need for robust process monitoring. Unlike oX-axis vision-based monitoring systems, which are constrained by line-of-sight and sensitive to geometric variation, acoustic sensing is flexible, low-cost and largely geometry-independent. In this study, acoustic emission (AE) signatures were systematically investigated in co-axial DED-LB/w of nano-treated Aluminium 7075 under three representative process regimes: Lack of Fusion, Conduction and Overheating. The findings reveal that the acoustic signal encodes both dripping-specific and process- and regime-specific features, which were then used to train supervised machine learning models for dripping detection and regime classification. The novelty of this work is an acoustic-emission-based, machine-learning-assisted monitoring framework for co-axial DED-LB/w, enabling in-process identification of dripping and the process regime with high accuracy. By providing geometry-independent sensing and real-time decision support, the framework enhances process stability and reduces unplanned interruptions, supporting wider industrial adoption of co-axial DED-LB/w technology.