Search Results (4,738)

Automatic License Plate Recognition (ALPR) in unconstrained traffic environments requires simultaneously addressing two fundamental challenges: reliable localization of small and degraded license plates and accurate decoding of visually ambiguous character sequences. This paper presents a hierarchical multi-stage framework that combines deep-learning-based detection, geometric normalization, dual-channel recognition, and structured post-correction to improve recognition robustness under diverse real-world conditions. A systematic ablation study involving five configurations (A0–A4) demonstrates the effectiveness of the proposed architecture across three benchmark datasets. On the UC3M-LP dataset, exact-match accuracy increases from 45.2% to 88.3%, while achieving 91.6% partial accuracy and a zero detection-miss rate. The framework further attains 95% exact-match accuracy on controlled European license plate crops and 93% on a large-scale custom dataset. In addition, we identify systematic evaluation artifacts in partially annotated benchmarks, showing that truncated ground-truth labels can underestimate genuine character-level improvements. The proposed framework supports multiple license plate formats through a configurable structural template library, and preliminary experiments on a small Arabic-script subset suggest potential extensibility without full model retraining. To ensure full reproducibility, all source code and evaluation resources are publicly released. Full article

Background and Objectives: Faricimab, the first bispecific antibody targeting VEGF-A and angiopoietin-2, has demonstrated durable efficacy in pivotal phase 3 trials for neovascular age-related macular degeneration (nAMD) and diabetic macular edema (DME). Real-world data on treatment-naïve patients managed with fixed-interval maintenance protocols, particularly from South-Eastern Europe, remain limited. This pilot study evaluated 24-month outcomes of intravitreal faricimab in treatment-naïve nAMD and DME, using a standardized four-injection loading phase followed by fixed every-16-week (Q16W) maintenance. Materials and Methods: This study conducted a retrospective, observational, single-center pilot cohort study of 20 consecutive treatment-naïve eyes (9 nAMD, 11 DME). All patients received four monthly loading injections followed by a fixed every-16-week (Q16W) maintenance schedule, supplemented by discretionary additional injections for residual or recurrent disease activity (215 injections total; mean 10.75 ± 0.79 per patient; range 9–12). Primary outcomes were changes in central foveal thickness (CFT) and best-corrected visual acuity (BCVA; Snellen lines with ETDRS letter equivalents) at months 4 and 24. Prespecified secondary analyses included bootstrap 95% confidence intervals, a linear mixed-effects model with a time × disease-group interaction, Bayesian credible intervals with weakly informative priors, false-discovery-rate (FDR) correction, and a minimum detectable effect-size analysis. Results: All 20 eyes completed 24-month follow-up. In nAMD, mean CFT decreased by 186.9 ± 71.9 µm (35.9%; bootstrap 95% CI 148.1–236.0; p < 0.001; d = 2.60), and BCVA improved by 3.89 ± 0.78 Snellen lines (~19 ETDRS letters; 95% CI 3.44–4.33; p < 0.001; d = 4.97). In DME, CFT decreased by 197.7 ± 65.7 µm (39.3%; 95% CI 162.5–237.3; p < 0.001; d = 3.01), and BCVA improved by 4.55 ± 1.04 lines (~23 ETDRS letters; 95% CI 4.00–5.09; p < 0.001; d = 4.39). All 20 eyes (100%) achieved ≥ 3 Snellen lines gain and ≥20% CFT reduction; 80% reached final BCVA ≥ 7 lines. A linear mixed-effects model showed a significant time effect (p < 0.001) but no time × group interaction (CFT p = 0.84; BCVA p = 0.51), indicating concordant trajectories across diseases. Bayesian analysis with weakly informative priors yielded posterior P(|d| > 0.8) ≥ 0.99 for all primary outcomes. After FDR correction, all pre-specified primary comparisons remained significant. The minimum detectable effect size with the realized sample sizes (Cohen’s d ≈ 0.66 combined, 1.07 nAMD, 0.94 DME at 80% power) was substantially below all observed effect sizes. No ocular or systemic adverse events were recorded. Conclusions: In this small, single-center, treatment-naïve pilot cohort, a fixed Q16W faricimab maintenance schedule with discretionary additional injections was associated with durable anatomical and functional improvements over 24 months in both nAMD and DME, with no adverse events recorded across 215 injections. Given the limited sample, these findings should be regarded as hypothesis-generating. The high responder rates likely reflect the cohort’s substantial baseline visual impairment (mean baseline BCVA ~20/120–20/200), which provides greater absolute capacity for measurable gain than in higher-acuity registration trial populations. These pilot data support fixed-interval faricimab as a logistically feasible candidate strategy in resource-constrained settings and should be confirmed in larger multicenter cohorts using standardized ETDRS acuity assessment. Full article

The introduction of respiratory multiplex PCR in the post-pandemic world has improved the detection of viral infections, whose clinical relevance is still being characterized. Patients with diabetes mellitus (DM) exhibit altered innate immune responses, yet the effect of concurrent viral infection on their inflammatory trajectory and clinical outcomes remains poorly characterized. This study examined whether diabetes is associated with a more pronounced inflammatory response, delayed resolution, and worse multi-organ outcomes during viral respiratory infections. A prospective, longitudinal cohort of 430 hospitalized adults (DM: n = 211; non-DM: n = 219) with PCR-confirmed viral respiratory infections was stratified into four groups by diabetes and co-infection status using a respiratory multiplex PCR panel. Serum IL-6, CRP, NLR, procalcitonin, and urea were measured at admission (Day 1) and at clinical stabilization (Day 6). All variables failed normality testing (Shapiro–Wilk p < 0.0001); non-parametric methods were applied. Receiver operating characteristic (ROC) analysis was used to identify candidate biomarker cutoffs for mortality prediction. SARS-CoV-2 was the predominant pathogen (29.1%). In an exploratory comparison limited by small subgroup sizes (n = 8 vs. n = 14), co-infected diabetic patients had higher baseline inflammatory markers than co-infected non-diabetic patients: median IL-6 32.87 vs. 6.20 pg/mL (Mann–Whitney p = 0.0006) and median CRP 103.83 vs. 23.03 mg/L (p = 0.0012). At the Day 6 checkpoint, co-infected diabetic survivors had higher IL-6 (12.01 vs. 6.13 pg/mL, p = 0.0183) and showed little within-group NLR change (Wilcoxon p = 0.2367); these Day 6 estimates are subject to survivor selection and should be interpreted accordingly. In-hospital mortality was 25.6% in diabetic vs. 3.7% in non-diabetic patients (p < 0.0001). Diabetic patients more frequently required orotracheal intubation (6.4% vs. 1.0%, p = 0.0207) and high-flow nasal oxygen (HFNO) support (7.9% vs. 1.8%, p = 0.0166). In an internal ROC analysis, baseline IL-6 showed the highest discriminatory performance for in-hospital mortality (AUC 0.812, 95% CI 0.772–0.848), with a candidate cutoff of > 55.78 pg/mL (sensitivity 71.0%, specificity 79.1%); IL-6 outperformed CRP (AUC 0.706, DeLong p = 0.0029) and NLR (AUC 0.656, DeLong p = 0.0001). As this cutoff was derived and evaluated in the same cohort, it is reported as exploratory and requires external validation. In this single-center cohort, diabetes was associated with a more pronounced baseline inflammatory profile, slower resolution of the neutrophil-to-lymphocyte ratio, and greater multi-organ involvement during viral respiratory infection, including in the small co-infected subgroup. In the full cohort, diabetes remained associated with higher mortality, IL-6, and CRP after adjustment for age, sex, and BMI; however, the small co-infected subgroups could not be adjusted, so those specific comparisons should be regarded as hypothesis-generating and need confirmation in larger, adequately powered multi-center cohorts. Full article

Energy Geographies has a unique position at the intersection of geospatial and social science, and it now faces a defining methodological development with the rapid rise in Geospatial Artificial Intelligence (GeoAI). This paper examines where GeoAI has and has not been applied within energy research through two bibliometric analyses using the Dimensions database. The first establishes an updated picture of energy geographies scholarship from 2020 to 2026, mapping the field’s current priorities and geographic distribution as a baseline for evaluating GeoAI’s role. The second conducts a bibliometric analysis of GeoAI-specific energy publications from 2020 to 2026, which reveals significant GeoAI Application Gaps: a heavy concentration in energy extraction and production research and in renewable energy siting and grid optimization, while energy transition, justice, and the energy problems of underrepresented regions remain substantially underserved. GeoAI energy research is also more geographically concentrated than the broader field, dominated by a small number of countries, raising questions about the applicability of these tools to the energy challenges facing the rest of the world. We argue that this gap reflects a pattern of problem selection as much as technological limitation, and that energy geographers are well positioned to redirect the development of this new field. We outline three directions for future research: developing Explainable GeoAI to ensure transparency and accountability, expanding geographic coverage to address data biases that favor a small set of well-resourced countries, and confronting the computational energy paradox of carbon-intensive AI applied to sustainability-oriented research. Full article

Treatment of non-small cell lung cancer (NSCLC) has been transformed by immunotherapy and targeted therapies. We aimed to characterize clinical features, treatment patterns, and healthcare resource use in patients with early and locally advanced NSCLC before incorporation of these therapies. This retrospective observational study included adults diagnosed with stage I–III NSCLC at four Spanish hospitals between 2014 and 2018, with follow-up until 2021, using artificial intelligence to extract data from electronic health records. A total of 951 patients were included (34.7% stage I, 16.7% stage II, 48.6% stage III), with a median age of 66 years and 31.9% female. Surgery was performed in 78.5% of stage I, 74.8% of stage II, and 35.5% of stage III patients. Among surgical patients, 62.5% received adjuvant chemo- and/or radiotherapy, 20.8% neoadjuvant therapy, and 15.7% both; among non-surgical patients, chemoradiotherapy was the most common treatment (50.4%). Beyond hospitalization, outpatient visits were the most frequently used healthcare resource. These findings provide a historical benchmark of NSCLC care before introduction of immunotherapy and targeted therapies in these settings, highlighting treatment variability and the need for earlier diagnosis, structured treatment pathways, and multidisciplinary management. Full article

Background: KRAS mutations are among the most common oncogenic drivers in non-small cell lung cancer (NSCLC) and are associated with substantial molecular and immunological heterogeneity. However, the clinicopathological associations and prognostic relevance of KRAS mutation subtypes and co-occurring genomic alterations in relation to PD-L1 expression and survival outcomes remain incompletely understood, particularly in the immunotherapy era. Methods: This retrospective single-center study included 93 KRAS-mutant NSCLC patients identified among 543 consecutively sequenced cases between March 2024 and March 2025. KRAS mutation subtypes, co-mutations involving TP53, STK11, and KEAP1, PD-L1 expression status, clinicopathological features, and survival outcomes were evaluated. Overall survival was assessed using Kaplan–Meier analysis and Cox proportional hazards regression models. Results: KRAS mutations were detected in 17.1% of NSCLC patients. G12C was the most frequent KRAS subtype (38.7%), followed by G12V (18.3%) and G12D (14.0%). Co-occurring mutations were identified in 73.1% of cases, most commonly involving TP53 (40.9%) and STK11 (33.3%). PD-L1 expression was negative in 48.4% of patients, low in 28.0%, and high in 23.7%. No significant association was identified between KRAS mutation subtype and PD-L1 expression (p = 0.663). STK11-mutated tumors demonstrated a trend toward lower PD-L1 expression levels compared with STK11 wild-type tumors. However, none of the molecular variables retained independent prognostic significance. Immunotherapy was associated with significantly prolonged overall survival (median OS: 24 vs. 7 months, p = 0.013) and remained independently associated with improved survival in multivariate analysis (HR: 0.376, 95% CI: 0.204–0.694, p = 0.002). Advanced-stage disease independently predicted worse survival outcomes (HR: 13.43, 95% CI: 1.81–99.79, p = 0.011). Conclusions: KRAS mutation subtypes and co-occurring genomic alterations demonstrated limited independent prognostic significance in this real-world NSCLC cohort. In contrast, immunotherapy was associated with improved overall survival in this retrospective cohort. These findings should be interpreted as observational and hypothesis-generating rather than evidence of predictive treatment benefit. Larger prospective studies integrating genomic and immune biomarkers are warranted. Full article

Background: Biologic therapies have transformed severe asthma management, but real-world evidence comparing phenotypes, lung function trajectories, and persistence across biologic classes remains limited. Objective: To characterize a real-world cohort of biologic-treated severe asthma patients, focusing on baseline phenotypes, longitudinal post-bronchodilator spirometry (including a spirometric surrogate suggestive of small airways involvement), and discontinuation/switching patterns. Methods: In this retrospective observational study at a tertiary referral center, adults with severe asthma treated with benralizumab, mepolizumab, omalizumab, or tezepelumab were included. Demographic, clinical, biomarker, and functional data were collected at baseline and follow-up. Post-bronchodilator FEV₁ and FEF25–75 (% predicted) were assessed at baseline, 6 months, 12 months, and 24–36 months when available. Longitudinal outcomes were analyzed using multivariable linear mixed-effects models; discontinuation and switching were recorded. Results: Eighty-seven patients were included (benralizumab n = 13, omalizumab n = 10, mepolizumab n = 30, tezepelumab n = 34), representing 10.9% of the clinic’s population. Most had long-standing disease, elevated body mass index, and a T2-high profile. Baseline characteristics were generally similar across groups, with expected differences in total IgE (p = 0.007) and blood eosinophils (p < 0.001). The primary endpoint (FEV₁ % predicted change from baseline to 12 months) showed adjusted mean changes of +12.46 (95% CI +1.63 to +19.29; p = 0.020) with benralizumab, +15.82 (+8.35 to +23.64; p < 0.001) with mepolizumab, +16.65 (+1.58 to +31.71; p < 0.001) with omalizumab, and +15.69 (+6.52 to +24.87; p = 0.030) with tezepelumab; trajectories differed by biologic class (time × biologic p = 0.019). Although the interaction term indicated heterogeneous temporal patterns, these adjusted findings should be interpreted as associative in the context of biomarker-driven treatment selection and not as evidence of comparative superiority of any biologic class. Discontinuation occurred in 15/87 (17.2%), with switching most commonly due to inadequate control. Conclusions: Real-world severe asthma patients demonstrate heterogeneous phenotypes and spirometric trajectories on biologics. Integrating biomarkers with longitudinal lung function monitoring, including small-airway spirometric surrogates, supports individualized management. Full article

Background/Objectives: Enhanced Recovery After Surgery (ERAS) protocols improved outcomes in colorectal surgery, but global implementation remains heterogeneous. In centers without structured ERAS programs, the lack of standardized discharge criteria could lead to conservative decisions and prolonged hospital stays. This study aimed to evaluate the feasibility and safety of early discharge (ED) on postoperative day (POD) 3 using the five Tavernier’s criteria in a real-world setting without formal ERAS pathways. Methods: This retrospective analysis of a prospectively maintained database included all consecutive adult patients undergoing elective laparoscopic colorectal resection between February 2025 and February 2026 at a high-volume tertiary center. Patients were stratified into the EARLY group (discharged on POD 3 upon fulfilling all five Tavernier criteria: C-reactive protein < 150 mg/L, temperature < 38 °C, passage of flatus, Visual Analogue Scale score < 5, and oral diet tolerance) and the STANDARD group (discharged after POD 3). The primary endpoint was the safety and negative predictive value (NPV) of the five-criteria bundle regarding 30-day complications. Results: Seventy-seven patients were included (EARLY: n = 44; STANDARD: n = 33). In the STANDARD group, the primary barriers to discharge were prolonged intravenous analgesic requirements (81.8%) and delayed bowel function (36.4%). The five-criteria bundle demonstrated an NPV of 84.1%, a sensitivity of 68.2%, and a specificity of 67.3% for identifying patients at low risk of complications. The overall 30-day complication rate was significantly lower in the EARLY group compared to the STANDARD group (15.9% vs. 45.5%; p = 0.010). No major complications (Clavien–Dindo ≥ III) occurred in the EARLY group compared to 6.1% in the STANDARD group. Conclusions: This exploratory feasibility analysis suggests that early discharge on POD 3 guided by the five Tavernier criteria is potentially safe and feasible in a real-world clinical setting without formal ERAS pathways. However, given the small sample size and inherent methodological biases, these findings remain preliminary, and larger prospective multi-center trials are strictly required to validate the safety and formal impact of this strategy. Full article

Background: Microcalcifications (MCs) are among the earliest mammographic signs of breast cancer, yet their detection remains challenging due to small size, low contrast, and dense breast tissue. This scoping review synthesizes AI-driven methods for MC detection in digital mammography, focusing on three dimensions: comparative performance of deep learning (DL) versus traditional methods, the clinical impact of explainable artificial intelligence (XAI), and the role of synthetic data in addressing dataset limitations. Methods: Following PRISMA-ScR guidelines, we systematically searched seven databases for studies published between January 2000 and January 2026. Of 366 initial records, 72 peer-reviewed studies were included in the final synthesis. Results: DL architectures, particularly convolutional neural networks (CNNs), have generally reported higher diagnostic performance (accuracy up to 99.71% and Area Under the Curve (AUC) up to 0.998) than traditional machine learning methods, although direct comparisons are hindered by heterogeneous datasets and evaluation protocols. XAI techniques have yet to undergo rigorous validation in real-world clinical settings, with very low certainty of evidence regarding their impact on radiologists’ trust or workflow integration. Synthetic data generation mitigates some data scarcity and privacy constraints but introduces artifacts (e.g., checkerboard patterns in 39–46% of cases) that limit clinical realism. Conclusions: DL offers substantial promise for MC detection, but translation to clinical practice requires robust XAI validation, higher-quality synthetic data, and prospective studies on diverse, longitudinal datasets. Full article

Background and Clinical Significance: Cutaneous squamous cell carcinoma (CSCC) is a common non-melanoma skin cancer, and while most cases are curable, a small proportion progresses to locally advanced or metastatic disease. As neoadjuvant immunotherapy with PD-1 inhibitors such as cemiplimab becomes more widely adopted, understanding real-world patterns of response remains essential. Case Presentation: We report a case of a man in his 50s with a large, locally advanced CSCC of the left hand in whom neoadjuvant cemiplimab was chosen to reduce tumor burden and preserve hand function when margin-negative resection was unlikely. The patient received four cycles of cemiplimab and demonstrated marked clinical improvement followed by complete pathological response at the primary site upon wide local excision. However, metastatic involvement of the epitrochlear and axillary lymph nodes was identified at surgery despite initial benign imaging. Postoperative PET/CT showed no additional disease, and the patient subsequently underwent axillary dissection and adjuvant cemiplimab with good functional recovery. Conclusions: This case highlights the potential for neoadjuvant cemiplimab to achieve substantial local tumor control and functional preservation while emphasizing the need for careful nodal assessment and ongoing surveillance in patients with very-high-risk CSCC. In cases where baseline cross-sectional staging is not performed, pre-existing occult nodal disease cannot be excluded. Full article

Biogenic or microbial methane (CH₄) emissions, believed to be the main driver of the recent surge in global atmospheric CH₄ emissions, have altered monitoring, measurement, and mitigation of fossil-fuel emissions. As of 1981, over 20% of the world’s natural gas reserves were biogenic in origin. Additional biogenic CH₄ reserves from coal have been discovered since 1981 mixed (40–80%) with thermogenic CH₄. Biogenic CH₄ accumulates up to 100% in coal reservoirs in the Powder River Basin (PRB), USA. Biogenic CH₄ is generated by microbial breakdown of fossil organic matter as an early-stage (primary) type during burial over geologic time and is rarely preserved. Also, biogenic CH₄ is generated as a late-stage (secondary) type from recent geologic to present times and is commonly preserved. Late-stage biogenic CH₄ is sustained by nutrients and microbes in meteoric/surface waters discharged into coal aquifers. Groundwater is pumped from wells in coal aquifers to desorb and produce CH₄ and dewater coal mines. The co-produced water with dissolved CH₄ is discharged into diverse surface aquatic systems. The emission factors (EFs) of co-produced water are 2.0522 × 10⁻⁹ Gg CH₄/gal of water in the PRB and 2.0694 × 10⁻³ Gg CH₄/well in the Black Warrior Basin, U.S. Accurate data on biogenic CH₄ emissions from coal sources is a major gap in the accounting of current global groundwater-driven CH₄ whose average flux is estimated to be 3.9 ± 6.2 mmol/m²/day or accounting for up to 70% of CH₄ emissions from surface aquatic systems. Biogenic CH₄ emissions from coal mining and coalbed gas extractions and related infrastructures are overlooked because the focus has been on coalmine methane (CMM) emissions. CMM data from ground-based measurements is highly variable and used by the Intergovernmental Panel on Climate Change three-tier system to estimate EFs for national inventories. However, 90% of CMM emissions are attributable to a small group of the most coal-consuming-and-producing countries but fails to capture other coal sources worldwide. This created gaps and “blind spots” in “unstructured” low-concentration, diffused biogenic CH₄ emission data. These key “blind spots” include sources from flooded, abandoned coal mines; coalbed methane (CBM) co-produced water with dissolved CH₄ and infrastructures/facilities; and groundwater drawdown from water withdrawals during coal mining and CBM extraction. Also, a critical “blind spot” is the mixing of biogenic CH₄ emissions from subsurface coals with biogenic CH₄ generated at the surface from wetlands, agriculture, and landfills/wastes, which grew 85% from 2008 to 2020. Limited understanding of the mixing of biogenic CH₄ from diverse sources and their contributions to global methane requires accurate attribution of overlapping isotopic signatures (δ¹³C_CH4 and δD). This paper addresses knowledge gaps in coalbed biogenic CH₄ emissions by a systematic review of the literature and specific study cases, which provided insights on key “blind spots” in their mitigation. Full article

Anomaly detection systems deployed in critical applications require not only high accuracy but also reliable uncertainty quantification and coverage guarantees. This paper is an empirical study—rather than a contribution of new conformal-prediction machinery—of class-conditional (Mondrian) conformal prediction for anomaly detection under extreme class imbalance, characterizing where standard conformal prediction fails and how class-conditional calibration restores valid coverage. Class-conditional conformal prediction constructs prediction sets that, under exchangeability, contain the true label with user-specified confidence (e.g., 90%), enabling systems to abstain on uncertain predictions. Unlike standard conformal prediction that fails catastrophically under extreme imbalance—achieving only 52.94% anomaly coverage at a 1:345 imbalance ratio—class-conditional calibration maintains 90.59% anomaly coverage by computing quantiles separately for each class. We apply the standard softmax-based nonconformity score $s=1-f_y\left(x\right)$ within each class, ensuring valid coverage for both normal and anomalous instances with coverage gaps ranging from 0.50% to 5.18% depending on dataset characteristics. Extensive experiments on three real-world datasets (Microsoft Azure KPI, Yahoo, NAB) demonstrate that the method achieves empirical coverage within 0.06–0.33% of theoretical targets at confidence levels $\alpha \le 0.05$; on the most imbalanced benchmark (Microsoft Azure KPI at a 1:345 ratio and $\alpha =0.10$), this corresponds to a 37.65 percentage point improvement in anomaly coverage over standard conformal prediction. We restate finite-sample coverage bounds and exchangeability conditions in the binary anomaly detection setting and validate them empirically through Monte Carlo simulation. Multi-model evaluation across XGBoost, Random Forest, and Neural Networks demonstrates the model-agnostic property of the framework, while also identifying conditions (poor base-classifier discrimination, small minority calibration sets) under which coverage may be marginally violated. Comparison with alternative uncertainty quantification methods (isotonic probability calibration, Monte Carlo dropout) shows that only conformal prediction provides formal guarantees while maintaining 90.59% anomaly coverage versus 76.47% and 84.71% for alternatives. The abstention mechanism identifies 34–66% of predictions as uncertain at high confidence levels (99%), enabling safety-critical systems to defer difficult cases to human experts while preserving baseline discrimination (ROC-AUC unchanged). Full article

The electrification of urban transport has made battery electric buses (BEBs) an important option for reducing carbon emissions and improving urban air quality. However, the high investment cost of charging infrastructure and the uncertainty in effective usable battery capacity at the day-ahead scheduling stage—caused by accumulated degradation, heterogeneous operating conditions, and imperfect state estimation—create major challenges for charging infrastructure siting and daily bus operations. This study proposes a joint optimization model for infrastructure siting and BEB charging scheduling, in which effective capacity uncertainty is handled using a distributionally robust optimization (DRO) framework. To solve the resulting mixed-integer nonlinear program efficiently, we develop a matheuristic decomposition method that integrates Adaptive Large Neighborhood Search (ALNS) with small gaps relative to a relaxation-based lower bound. Computational experiments based on real-world bus route data indicate that the proposed framework obtains high-quality solutions with small gaps relative to a relaxation-based lower bound, performs better than representative benchmark heuristics, and scales well to large instances. Full article

Small municipalities often lack the resources and infrastructure necessary to implement advanced digital twin solutions commonly adopted in larger cities or industries. This study addresses the challenge of designing a replicable and interoperable local digital twin architecture specifically suited for low-infrastructure environments. A gap in the current literature and practice is identified: most digital twin implementations are domain-specific, resource-intensive, or proprietary, which limits their applicability in low-infrastructure contexts such as small rural areas. To address this issue, a requirement-driven architecture based on open standards and minimal-footprint, edge-based technologies is proposed. The approach is validated through real-world implementation in Codogno, Italy, with subsequent replication in Varna, Bulgaria, and Lausanne, Switzerland. The findings indicate that the proposed architecture can be deployed with minimal local infrastructure while maintaining interoperability with existing systems and enabling scalability to larger contexts. Interoperability is achieved through standardized data models and APIs, while replicability is ensured by a modular design utilizing open-source components. These contributions offer a practical blueprint for small municipalities to develop local digital twins, thereby supporting digital transformation at the community level. Full article

Accurate and robust weed detection is a critical prerequisite for precision agriculture and site-specific weed management. However, real-world agricultural environments pose significant challenges to existing object detectors due to severe illumination variability, high inter-class similarity between crops and weeds, and the prevalence of small and occluded targets at early growth stages. To address these challenges, this paper proposes DSD-YOLOv11, a domain-adaptive and structurally refined detection framework tailored for complex field scenarios. Specifically, a physics-based data augmentation strategy is first introduced to simulate realistic illumination conditions and soil background variations, effectively broadening the training distribution without increasing model complexity. In addition, a lightweight Feature Enhancement Module (FEM) is selectively injected at the P3 detection layer, where high-resolution features are preserved. The FEM integrates a SpatialAttentionLite mechanism with a projection-based feature alignment strategy, enabling precise enhancement of fine-grained spatial cues while maintaining compatibility with pre-trained backbones. An epoch-aware alpha controller is further designed to ensure stable optimization by gradually activating the enhancement pathway during training. Extensive experiments on a real-world agricultural weed dataset demonstrate that the proposed method consistently outperforms baseline YOLOv11 models across multiple evaluation metrics. Notably, DSD-YOLOv11 achieves an absolute mAP@50 improvement of +12.73 percentage points over the native baseline without data augmentation (reaching 87.14%, where the physics-based augmentation contributes +7.94 percentage points and the FEM module contributes an additional +4.79 percentage points over the augmented YOLO11n baseline), while operating at 84.2 FPS on a desktop GPU (NVIDIA RTX 4090; NVIDIA Corporation, Santa Clara, CA, USA) and 7.2 FPS on an edge computing platform (NVIDIA Jetson Nano; NVIDIA Corporation, Santa Clara, CA, USA) with only marginal parameter increases. These results indicate that the proposed framework provides an effective and efficient solution for weed detection in unstructured agricultural environments. Full article