To review recent advances in the use of artificial intelligence (AI) to address shortcomings in assessing and improving surgical performance/training by automating surgical skills assessment and feedback. We searched PubMed for studies published between 2015 and 2025 pertaining to AI for surgical training. Search terms included 'artificial intelligence or 'machine learning' or 'deep learning' and 'surgical feedback' or 'surgical training' or 'surgical skill'. Articles were identified with special attention given to those published in the last 5 years with a focus on AI for surgical skill assessment or feedback. Artificial intelligence has been used to successfully automate surgical skill assessment across a variety of surgical disciplines via approaches such as kinematics, sabermetrics, computer vision, and gesture analysis. Many of these studies have developed AI models capable of a binary classification of skill (novice vs expert), which demonstrate concordance when verified against ground truths from human raters. Based on these skills assessments, AI approaches may be further leveraged to generate automatic feedback, which has proven effective in improving surgeon performance metrics, particularly for underperformers. AI has also shown utility in categorising and analysing the content and impact of live surgical feedback, enabling more efficient analysis of how feedback can be best delivered to trainees. Artificial intelligence is a promising tool for augmenting surgical training and improving the objectivity and scalability of surgical skill assessment and feedback. To date, AI models are adept at detecting relatively large differences in surgical performance and providing rudimentary feedback. Further work is required to create models capable of doing more fine-tuned skill assessments and generating more detailed, constructive feedback.

This study addresses the problem of robust actuator fault estimation for a class of critical linear continuous-time systems subject to state time-varying delays, external disturbances, and actuator faults. A learning observer is proposed to achieve the challenging task of simultaneously estimating both the system states and actuator faults, irrespective of whether the faults are constant or time-varying. A key theoretical contribution is the derivation of a less conservative delay-dependent condition for the existence of the proposed learning observer, which is expressed in terms of linear matrix inequalities (LMIs). The H∞ performance index is employed to attenuate the effects of disturbances to a prescribed level. The efficacy of the proposed strategy is rigorously validated through three illustrative examples, including quantitative performance metrics and a comparative analysis with existing methods.

To enhance the accommodation capability and operational flexibility of renewable energy systems, address the insufficient architectural integration of existing ammonia-based energy systems, and overcome the limitations of current optimization algorithms in tackling complex nonlinear multi-objective problems, this paper proposes a synergistic integrated energy system with liquid ammonia as the central hub. The system integrates multi-energy flows encompassing electricity, hydrogen, ammonia, and heat, leveraging ammonia fuel cell power generation, ammonia cracking, and ammonia-blended gas turbines for both electricity and heat production. A bi-level optimization model is formulated, coupling upper-layer multi-objective capacity planning with lower-layer stochastic chance-constrained scheduling. To solve this model, a hybrid algorithm, designated as LSDBO-WOA, is developed by integrating an improved dung beetle optimizer (LSDBO) with the whale optimization algorithm (WOA). Case study results demonstrate that the proposed algorithm achieves markedly superior convergence performance compared to benchmark algorithms such as non-dominated sorting genetic algorithm II (NSGA-II), with an improvement of approximately 18.6% in comprehensive performance metrics. Furthermore, the proposed electricity-hydrogen-ammonia-heat system attains an overall energy efficiency exceeding 97.66% and reduces carbon emissions by 7.3% relative to the original system without ammonia integration.

Structured Interdisciplinary Rounds (SIDRs) aim to improve communication, care coordination, and discharge planning by bringing multidisciplinary teams together for structured bedside discussions. Although widely promoted, few multisite evaluations have assessed their association with key hospital performance metrics compared with traditional rounding models. To evaluate the association of SIDRs with hospital efficiency, safety, and patient experience across a multi‑hospital health system. Retrospective cohort study comparing SIDR and traditional care units across four hospitals. A total of 11,334 inpatient discharges between July 1, 2023, and October 31, 2024. Daily structured interdisciplinary rounds conducted by physicians, nurses, case managers, pharmacists, and rehabilitation staff. Length of stay (LOS), observed‑to‑expected LOS (O/E LOS), case mix index (CMI)‑adjusted LOS, 30‑day readmissions, safety outcomes (falls, pressure injuries, medication errors per 1000 patient‑days), patient experience (HCAHPS communication and care‑transition domains), and complaints per 1000 patient‑days. SIDR units had lower O/E LOS compared with traditional units (1.35 vs 1.50; Δ - 0.15, 95% CI - 0.22 to - 0.05; p = 0.004). Unadjusted LOS was higher in SIDR units, whereas CMI‑adjusted LOS favored SIDR among moderate‑ and high‑complexity patients. Thirty‑day readmissions and patient‑experience scores did not differ significantly. Safety event rates were low in both groups, with no significant differences, although medication‑error reporting was likely under‑captured due to voluntary reporting systems. Unit‑level sensitivity analyses demonstrated site‑level heterogeneity but were directionally consistent with patient‑level findings, with risk‑adjusted advantages for SIDR most pronounced at one hospital. SIDRs were associated with lower risk‑adjusted LOS without differences in readmissions, safety events, or patient‑experience scores. Benefits were greatest among higher‑complexity patients, suggesting that structured interdisciplinary communication may be particularly impactful for patients requiring intensive coordination. Further research should incorporate broader safety indicators, process‑of‑care measures, and patient‑reported experience tools to more fully characterize the effects of SIDRs across diverse inpatient environments.

High-dimensional data classification remains challenging for machine learning models due to sparsity and overfitting caused by the ‘curse of dimensionality‘. As the number of features increases, data points become sparse, hindering generalization in classification and leading to higher computational costs and reduced accuracy. To address these issues, we propose an ensemble classifier based on random subspaces implemented in the Spark framework. The proposed framework comprises three key stages. First, the high-dimensional data is normalised through min-max normalisation. Second, the master node partitions the data by using improved deep fuzzy clustering (IDFC). In contrast, the slave node applies support vector machine-modified recursive feature elimination (SVM-MRFE) for efficient feature selection, followed by feature fusion. Finally, we introduced an improved subspace-based ensemble classifier (ISSBEC) that comprises a feature-fusion-based random subspace (FF-RSS), mixed-space enhancement (MSE), and multiple base classifiers. The efficacy of the ISSBEC classifier was evaluated using a set of performance metrics and compared with state-of-the-art methods. Experimental results demonstrate that the proposed approach improves both accuracy and robustness, offering a scalable solution to the limitations of high-dimensional datasets.

The effective integration of artificial intelligence (AI) systems into clinical medicine depends on comprehensive and transparent performance evaluation; however, the lack of standardized and widely accepted metrics poses challenges for reproducibility and model adoption. A comprehensive, machine-interpretable framework is presented to formalize the nomenclature and descriptions of 207 graphical, matrix, and scalar metrics used to measure AI model performance. The metrics taxonomy, developed as part of the Radiology Ontology of AI Datasets, Models and Projects (ROADMAP), provides a logically structured representation that captures the semantics of AI evaluation metrics, supports reasoning over metric classes, and enables automated completeness checks for AI model reporting. For each metric, the taxonomy incorporates a definition and citations to authoritative reference sources; where applicable, the taxonomy also includes synonyms, abbreviations, alternate language forms, mathematical formulae, and numerical bounds. The taxonomy supports evaluation of models operating on structured data, medical images, audio signals, and/or unstructured text. Logical axioms link each metric to one or more of 18 AI model performance criteria, including classification, calibration, image segmentation, and text analysis. By harmonizing terminology and enabling structured queries, ROADMAP's taxonomy of AI performance metrics facilitates model comparison, bias detection, and selection of appropriate evaluation methods across diverse datasets and clinical tasks. © RSNA, 2026 See also accompanying Special Report on ROADMAP ontology.

Given the growing concern over the operational safety and long-term reliability of lithium-ion batteries, the accurate assessment of battery state of health (SOH) is of paramount importance. With the aim of elevating the SOH estimation exactitude and remedying the model degradation induced by data paucity, this paper proposes an SOH estimation method that integrates a data-augmentation strategy with a Long Short-Term Memory (LSTM)-iTransformer model. Specifically, multiple health characteristic factors characterizing the aging behavior are first extracted from the battery charge–discharge curves and incremental capacity (IC) curves, and the features that are highly correlated with the SOH are screened by a Pearson correlation coefficient analysis. Subsequently, the data augmentation technique is used to extend the degradation sample set. The LSTM-iTransformer model is trained based on the extended samples and evaluated on multiple performance metrics. A comparative analysis reveals a marked enhancement in predictive accuracy achieved by this method over the baseline model trained with the initial data, which validates the effectiveness of the data augmentation strategy in improving the performance of SOH estimation models. Additionally, in scenarios characterized by abundant data availability, the direct application of this model facilitates enhanced predictive precision.

Flexible endoscopic suturing tools are complex and may have a long learning curve. This porcine study evaluated the safety and performance of a simplified suturing system compared with a commercially available device for the repair of gastrointestinal mucosal defects. This IACUC-approved study included four healthy swine. A total of ten defects (six in the stomach and four in the rectosigmoid colon) were created in each animal. Defects were randomly assigned to closure with either the novel or the commercially available system using a therapeutic gastroscope. Technical success was defined as mucosal closure of the defect with the inability to visualize any significant portion of the resection bed. Additional performance metrics included procedure time and device ease of use (assessed using the NASA Task Load Index [TLI]) and adverse events. No adverse events were reported post-procedurally for any of the test animals. The proportion of target resection sites achieving technical success was 100% in both treatment groups. The mean SimpleStitch NASA-TLI score was lower compared to the OverStitch device. Closure times were similar between the two devices. Histological assessment scores indicated expected healing response without evidence of perforation, leakage, or abscess formation. A novel full-thickness suturing system safely and effectively closed mucosal defects. Lower NASA-TLI scores suggest that the novel suturing device may offer simpler, less demanding use compared to the predicate device, potentially reducing the learning curve for endoscopic suturing procedures.

Interpretable evaluation of physiological signals for biometric identification.

Modern supply chain management faces unprecedented challenges in demand forecasting and inventory optimization due to increasing market volatility, consumer behavior complexity, and global disruptions. Deep learning (DL) has emerged as a transformative approach that addresses these challenges by capturing complex nonlinear patterns in demand data and optimizing inventory decisions across multiple echelons. This review examines the current state of DL methods applied to demand forecasting and inventory optimization in supply chains. Recurrent neural networks (RNNs), long short-term memory (LSTM) networks, convolutional neural networks (CNNs), and transformer-based architectures have demonstrated superior performance compared to traditional statistical methods. The integration of DL with reinforcement learning (RL) has enabled adaptive inventory policies that respond dynamically to changing market conditions. Graph neural networks (GNNs) have proven effective in capturing network dependencies across complex supply chain structures. Despite these advances, challenges remain in model interpretability, data quality requirements, computational complexity, and real-time implementation. This paper provides a comprehensive analysis of DL architectures, hybrid approaches, performance metrics, and practical applications while identifying critical research gaps and future directions for advancing intelligent supply chain management systems.

Abstract Sequential filtering and spatial inverse problems assimilate data points distributed either temporally (in the case of filtering) or spatially (in the case of spatial inverse problems). Sometimes it is possible to choose the position of these data points (which we call sensors here) in advance, with the goal of maximising the expected information gain (or a different metric of performance) from future data, and this leads to an Optimal Experimental Design (OED) problem. Here we revisit an interpretation of optimising sensor placement as an integration with respect to a general probability measure $$\xi $$ ξ . This generalises the problem of discrete-time sensor placement (which corresponds to the special case where the probability measure is a mixture of Diracs) to an infinite-dimensional, but mathematically more well-behaved setting. We focus on the continuous-time stochastic filtering setting, whose solution is governed by the Zakai equation. We derive an expression for the Fréchet derivative of a general OED utility functional, the key to which is an adjoint (backwards in time) differential equation. This paves the way for utilising new gradient-based methods for solving the corresponding optimisation problem, as a potentially more efficient alternative to (semi-)discrete optimisation methods, e.g. based on greedy insertion and deletion of sensor placements.

To demonstrate feasibility of the MindCap mobile, digital application for objective and repeated assessment of cognitive function in cancer survivors and assess its sensitivity to cognitive improvement. 57 adults with stage I-III solid tumors or lymphoma who reported cognitive issues six months to five years postchemotherapy. Participants completed three phases of MindCap testing sessions over 14 weeks. Pre-/postintervention data were collected for time and number of MindCap sessions completed for each phase, adherence to required testing frequency, participant satisfaction, self-report, and MindCap metrics for cognitive domain-specific performance (executive function, attention, memory, verbal ability, visual-spatial ability, and psychomotor function). Feasibility was demonstrated for recruitment and adherence to MindCap use. Self-report and MindCap metrics correlated positively. MindCap performance metrics were sensitive to improvements in multiple cognitive domains. Post facto analyses indicated a potential MindCap dose effect for cognitive function improvement. Future research comparing MindCap metrics to standard neurocognitive testing and investigation for potential dose effect are warranted.

Group IV–VI monochalcogenides have attracted considerable interest for polarization-sensitive photodetection due to their pronounced chemical stability and photoelectric anisotropy. However, scaling high-quality single-crystalline films over large areas remains challenging, limiting the active device area and ultimately the performance metrics of linear-polarization photodetectors. Furthermore, the origin and gate-tunability of valley-selective photoresponses in these materials are not fully understood. Herein, sub-centimeter-scale single-crystal GeS films with high lattice uniformity are successfully synthesized. Photodetectors based on these films exhibit strong wavelength-dependent linear dichroism, reaching a maximum polarization ratio of 2.7 at 670 nm. Significantly, a photoinduced bilateral Schottky-barrier lowering effect is identified, which provides an internal photo-gain, leading to the higher responsivity of 17.37 mA/W and detectivity of 2.77 × 1010 Jones at 405 nm. A gate voltage-modulated rotation of the polarization-sensitive photocurrent axis is observed, revealing a complex interplay between intrinsic anisotropy and external-field effects. Theoretical analysis indicates that the intrinsic dichroism originates from two distinct in-plane polarized valleys, while the gate-induced angular rotation is primarily driven by an anisotropic Pauli-blocking mechanism. These findings not only demonstrate a viable route for wafer-scale preparation of single-crystal GeS films but also provide fundamental insights into the electric modulation of linear dichroism, advancing the development of high-performance polarized photonic devices.

Abstract Beef production systems have been evolving for more than a millennium with a primary focus on increasing efficiency. Recently, environmental, social and governance (ESG) concerns have become another key focus due to societal, consumer and regulatory pressures. We identify the temporal evolution of various beef production systems. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we assessed the ESG impacts of eight systems across five domains: natural capital; social capital; human capital; produced capital; and governance. This analysis includes 98 articles. Additionally, we conducted a strengths, weaknesses, opportunities and threats (SWOT) analysis. Regenerative grazing ranks high while landless industrialized system ranks low on all performance metrics. Silvopasture and pasture-based systems excel in community engagement. Biodynamic farming, certified organic and regenerative systems support livelihoods and demonstrate robust governance structures with active stakeholder engagement. Landless industrialized systems demonstrate strong corporate governance. This study provides policymakers with insights on promoting sustainable and ethical beef production practices.

Kidney stones are a prevalent urological condition with significant global burden, often diagnosed using ultrasound (US) as a first-line modality despite its limitations in sensitivity and operator dependency. Artificial intelligence (AI) and deep learning (DL) algorithms have shown promise in enhancing US-based kidney stone applications, including detection, classification, complication prediction, and procedural guidance, but evidence remains heterogeneous. To systematically review and synthesize the applications of AI and DL algorithms in US-based kidney stone detection, classification, prediction of complications/outcomes, and procedural guidance. This systematic review followed PRISMA guidelines (PROSPERO: CRD420251247650). Databases including PubMed, Embase, Scopus, and others were searched from inception without language restrictions. Eligible studies were original peer-reviewed articles evaluating AI/DL in US for kidney stone diagnostics against reference standards like CT or surgical findings. Two reviewers independently screened, extracted data, and assessed quality using QUADAS-2 with AI extensions. From 1,285 records, 9 studies were included after exclusions. These encompassed DL for image detection/segmentation (n = 3), predictive modeling for complications/outcomes (n = 4), and procedural guidance (n = 2). Methodologies included CNN variants and ML ensembles. Performance metrics were high, with accuracies up to 96.54%, AUCs > 0.90 for predictions, and improved procedural outcomes. Risk of bias was low in most studies (5/9), with some concerns in others. Heterogeneity in datasets and validation limited meta-analysis. AI and DL algorithms demonstrate high diagnostic accuracy and clinical utility in enhancing US for kidney stone management, with stratification by application type revealing high performance across tasks, addressing traditional limitations. However, methodological variability and low to very low certainty of evidence (per GRADE) necessitate standardized external validation and multimodal integration for broader adoption.

Flood inundation mapping (FIM) is essential in disaster risk management, infrastructure planning, and climate adaptation. Traditional hydrodynamic models, such as the Hydrologic Engineering Center’s River Analysis System (HEC-RAS) and LISFLOOD-Floodplain (LISFLOOD-FP), provide physically interpretable flood simulations but are often data- and computation-intensive and difficult to scale across regions. In recent years, machine learning (ML) and deep learning (DL) approaches have emerged as data-driven alternatives that leverage remote sensing observations, digital elevation models (DEMs), and hydro-climatic datasets to enable scalable and near-real-time flood mapping. Our review synthesizes recent advances in ML-based flood inundation mapping, categorizing methods into traditional machine learning techniques (e.g., Random Forest (RF), Support Vector Machines (SVM), Gradient Boosting (GB)), deep learning architectures (e.g., Convolutional Neural Networks (CNNs), U-Net, Long Short-Term Memory networks (LSTM)), and emerging hybrid and physics-informed frameworks. We evaluate model performance across flood extent and flood depth estimation tasks, highlighting strengths, limitations, and common benchmarking practices reported in the literature. The review identifies key challenges related to model interpretability, data bias, transferability, and regulatory acceptance, and highlights recent progress in explainable artificial intelligence (XAI), uncertainty-aware modeling, and physics-informed learning as pathways toward operational adoption. By unifying terminology, performance metrics, and methodological comparisons, this review provides a coherent framework for advancing trustworthy, scalable, and decision-relevant flood inundation mapping under increasing climate-driven flood risk.

To compete with fossil fuels, biofuels produced from renewable waste biomass must be cost-effective, adaptable to existing heat and power infrastructure, and possess desirable fuel properties and performance metrics matching those of fossil fuels, while having a much lower carbon footprint. However, handling and processing biowastes in thermochemical biorefineries is challenging owing to their high moisture content, low bulk density, poor grindability, low calorific value, and heterogeneous physicochemical properties. Torrefaction has emerged as an effective thermochemical technology for upgrading biowastes into torrefied biomass, which exhibits improved, homogeneous physicochemical properties, including higher calorific value, higher bulk density, better grindability, and hydrophobicity. This review synthesizes the current state of research on torrefaction, with particular emphasis on process parameters, reactor designs, commercial-scale implementations, and an analysis of its strengths, weaknesses, opportunities, and threats. The comparative advantages and limitations of different torrefaction reactors are highlighted, emphasizing how each reactor’s characteristics determine its suitability for specific circumstances and operating conditions. This article also considers the technical and economic challenges associated with scaling up torrefaction. The discussion on specific case studies on techno-economic analysis of torrefaction outlines the key barriers and provides incentives for researchers to consider when upscaling the technology. The strengths, weaknesses, opportunities, and threat analysis offers strategic insights for policymakers and industry stakeholders into possible actions to support torrefaction and its upscaling.

ABSTRACTBackground and AimCervical cancer is a major public health issue in Botswana, driven by high HIV prevalence and limited screening access. Molecular detection of high‐risk human papillomavirus (hrHPV) is crucial for prevention and early diagnosis. The aim of this narrative review is to synthesize studies evaluating seven HPV detection platforms used in Botswana, focusing on diagnostic performance, feasibility, and scalability.MethodsA comprehensive literature search was conducted across PubMed, PMC, Google Scholar, and regional repositories up to July 2025 to identify studies evaluating HPV detection platforms in Botswana. Eligible studies included those assessing molecular HPV assays or genotyping tools in Botswana populations, reporting diagnostic performance or feasibility data. Data extracted covered study population characteristics, sample types, detection platforms, hrHPV prevalence, and performance metrics. Findings were qualitatively synthesized to compare platforms by sample type and application, as heterogeneity precluded meta‐analysis.ResultsThe Cepheid Xpert HPV assay, with near‐point‐of‐care capability, self‐sampling suitability, and high sensitivity (96%) and specificity (90%), is well‐suited for national programs despite costs of $15–$25 per test. AmpFire, costing $10–$15 per test, requires minimal equipment, offers good field feasibility (92% sensitivity, 88% specificity), and is promising for community deployment. Laboratory‐based assays like Roche Linear Array and Abbott RealTime deliver broader genotyping and > 93% sensitivity but require centralized infrastructure. PathoChip microarray detects novel HPV types yet is cost‐prohibitive (> $150/test). Double‐nested PCR aids retrospective formalin‐fixed paraffin‐embedded (FFPE) analysis but lacks practicality. These platforms collectively enhance cervical cancer prevention strategies in Botswana.ConclusionsThis review identifies feasible HPV detection platforms for Botswana's cervical cancer control. Cepheid Xpert and AmpFire assays offer accurate, affordable, and field‐ready options for early diagnosis and wider screening access. Integrating these molecular tools into existing HIV and reproductive health services could enhance prevention, early detection, and progress toward WHO cervical cancer elimination goals.

Purpose Law enforcement agencies in the United States are relying on state fusion centers for intelligence to develop actionable, data-driven reports that increase efficiency and improve investigations in crime prevention and homeland security. This study assesses the extent to which artificial intelligence and machine learning (AI/ML) are increasingly shaping intelligence operations in law enforcement and the functions of state fusion centers in supporting intelligence-led policing (ILP). The reliance and integration of AI/ML is improving analytic accuracy, situational awareness and information and data sharing and collaboration among law enforcement and homeland security agencies. This study examines the state of contemporary academic literature, assesses AI/ML applications used in law enforcement and builds a conceptual and theoretical framework centered on ILP policing. It also relies on empirical data, case study applications, and DHS assessments to explore the degree to which AI-driven processes and analytics enhance criminal intelligence, investigative efficiencies, situational awareness and predictive policing. The analysis, while focusing on the opportunities and challenges of using AI/ML tools in law enforcement, also highlights the need for ethical governance, transparency and accountability when relying on advanced technologies for crime prevention and policing. Design/methodology/approach This study utilizes qualitative methods, including a thematic content analysis of government and think tank/practitioner reports as well as academic literature on the benefits, costs and ethical factors regarding variations in the implementation of AI/ML tools for law enforcement intelligence products and resource allocation. For cross-validation of operational outcomes, it examines publicly available information in the DHS Fusion Center Annual Assessment, Bureau of Justice Statistics, LEMAS and RAND Corporation assessments of intelligence-led policing. Findings Qualitative Results Federal and state sources report fusion centers and law enforcement agencies integrating advanced analytic and ML-enabled tools into each step in the criminal intelligence lifecycle process. However, ethical and structural challenges limit and constrain technology-driven narratives in fusion centers. Given these challenges, there is a consistent qualitative and thematic pattern: state fusion centers now function as criminal intelligence analytic hubs or resources that leverage the most contemporary analytic and data-driven tools for criminal intelligence and law enforcement investigations. Interrelated themes describe AI/ML technologies in terms of shaping, constraining and complicating the intelligence lifecycle in fusion centers and law enforcement operations. Seven specific themes emerged from latent coding are illustrated in the chart. Research limitations/implications There are limitations on the collection of quantitative data since DHS, leading think tanks and NGOs do not disclose specific figures on the proportion of AI/ML tools. The DHS Fusion Center Annual Assessment process monitors technology adoption and the growth of analytic capabilities throughout the national network of fusion centers; however, the specific quantitative statistics are not disclosed in public summaries (DHS, 2024). Second, publicly available data and information constitute the bulk of empirical sources. Consequently, this study relies primarily on qualitative narrative reporting, not quantitative performance metrics. Third, publication bias is likely present in industry and government sources as these reports may provide overly optimistic observations and conclusions while overlooking ethical dilemmas, failures and challenges. Moreover, qualitative thematic analysis could reflect broader structural narratives as opposed to empirical outcomes. Finally, since AI/ML adoption varies across fusion centers and according to technology levels, qualitative themes identified in this study must be read as representative patterns and not as universally generalizable. Originality/value Fusion center utilization of AI/ML technologies is as much an operational tool as it is a policy, governance and ethical challenge. Successful and professional use in support of law enforcement is about placing technological innovations firmly within institutional accountability and constitutional guardrails. On the one hand, AI/ML tools are enhancing analytical intelligence production by accelerating analytic workflows, predictive modeling and expanding data/information integration capabilities. AI/ML are extending ILP concepts by offering improvements in situational awareness and threat identification and operational efficiencies. On the other, substantial constraints hinder responsible use of these technologies. In the absence of standardized oversight frameworks, data-quality issues, algorithmic bias and the lack of professional development, workforce capacity and critical skills on the part of fusion center analysts will cancel the benefits of these tools.

The increasing use of artificial intelligence (AI) in smart environments, industrial automation, healthcare monitoring, and Internet of Things (IoT) networks has increased the computational requirements and power consumption of cloud computing infrastructure. The use of cloud computing for AI processing faces challenges such as high latency, high bandwidth consumption, privacy concerns, and environmental emissions due to the large number of data centers. This work introduces a sustainable and low-power AI model based on the smart edge architecture․ It includes energy-efficient machine learning algorithms deployed at the edge for making smart decisions․ This model achieves optimal performance using lightweight model deployment, dynamic resource allocation, and hardware-aware optimizations such as model quantization and hardware pruning techniques․ A modular architecture is proposed that increases efficiency by focusing on data acquisition, edge processing, AI inference, energy management, and cloud synchronization․ Performance metrics include latency, energy consumption, compute efficiency, and inference quality․ The latter, in particular, is important for providing efficient AI capabilities on devices with limited resources․ Results of experiments show that local processing reduces network and energy use compared to processing in the cloud․ The proposed framework enables scalable and efficient deployment of AI while minimizing the environmental impact and maintaining performance to support sustainable computing․