Search Results (19,250)

Background: Periplaneta americana residue (PAR) has potential as a sustainable insect-derived feed ingredient. This study evaluated the effects of 3% dietary PAR on growth performance and ovarian function in finishing pigs and screened potential bioactive peptides. Methods: Thirty finishing pigs were assigned to a control diet or a diet containing 3% PAR. Growth traits, serum oxidative and inflammatory indicators, and ovarian histomorphology were assessed in vivo. Multi-omics profiling was integrated with quantitative real-time PCR, Western blotting, and immunofluorescence assays. The properties of selected peptides and their potential interactions with candidate proteins were assessed separately using in silico prediction, molecular docking, and molecular dynamics simulations. Results: PAR supplementation increased average daily gain and the ovary-to-body weight ratio and decreased serum malondialdehyde levels and follicular atresia, without significantly affecting feed intake, feed conversion ratio, or inflammatory cytokines. Multi-omics profiling associated PAR supplementation with changes in ovarian metabolism, antioxidant defense, immune regulation, and phosphorylation-related signaling. MGST2 and PPM1K were identified as candidate proteins, and their expression patterns were supported experimentally. Serum peptidomics identified GVSEIQQ as having higher abundance in the PAR group than in the control group, whereas GVEEHETL and PGIPT were detected only in the PAR group. In silico analyses predicted that these peptides were non-toxic, non-allergenic, and potentially antioxidant; these predictions require experimental validation. Molecular docking and molecular dynamics simulations suggested potential interactions of GVSEIQQ with MGST2 and GVEEHETL with PPM1K. However, these computational findings do not establish direct peptide–protein binding or functional effects in vivo. Conclusions: PAR shows potential as an alternative feed ingredient that may partially replace soybean meal while supporting growth and ovarian condition in finishing pigs. Further biochemical and in vivo studies are required to confirm the biological functions and mechanisms of the identified peptides. Full article

Osteoarthritis (OA) is a chronic inflammatory and degenerative joint disease that commonly affects the aging population. This study was designed to decipher gene–metabolite regulatory networks driving OA progression via combined computational prediction and experimental validation. Differential gene expression analysis, weighted gene coexpression network analysis, and machine learning algorithms were integrated to screen for key regulatory genes. To infer causal associations between these candidates and OA susceptibility, we performed Mendelian randomization (MR) using cis-expression quantitative trait loci (cis-eQTL) as genetic instruments and identified significant associations between TNFRSF10B and OA. In parallel, single-nucleus RNA sequencing (snRNA-seq) was used to map the cell-type-specific expression profile of TNFRSF10B, and its expression was validated in human knee synovial tissues. The protective effect of TNFRSF10B on OA was statistically mediated in part by the alpha-tocopherol-to-sulfate ratio, as inferred from Mendelian randomization analysis, with the indirect effect accounting for 16.85% (95% CI: 2.96–30.74%) of the total protective effect. snRNA-seq data revealed that TNFRSF10B was widely expressed across synovial cell populations but reduced in OA samples. Collectively, this study identified TNFRSF10B as a candidate protective factor in OA, supported by genetic evidence from MR analysis and reduced expression in OA synovial tissues, offering statistical evidence for gene–metabolite interplay in OA pathogenesis. Full article

Background: Kinesins are ATP-dependent molecular motors that mediate intracellular transport, cytoskeleton remodeling, and abiotic stress responses in plants. The KIN gene family remains poorly characterized in Prunus mume. This study aimed to explore the evolutionary features and cold-responsive functions of the PmKIN gene family. Methods: A systematic genome‑wide analysis was performed in P. mume. We performed phylogenetic classification, conserved domain detection, gene duplication and selection pressure analysis, cis-element prediction, tissue expression profiling, cold stress expression assay, and protein interaction network prediction. Results: Fifty PmKIN family members were defined and grouped into 10 subfamilies. K14 subfamily contained the largest number of members. All members harbor conserved motor domains, and segmental duplication drove the expansion of this gene family, which was overall constrained by purifying selection. The PmKIN homologous genes were highly conserved among Rosaceae species, including Prunus persica, Prunus armeniaca, Prunus avium, and Malus domestica. Our promoter analysis detected abundant cis-elements for light, hormone, cold, and drought signals in PmKIN genes, especially in the K14 and K7 subfamilies. These genes displayed tissue-biased expression, with roots and stems showing much higher transcript levels than fruits. Under low-temperature stress, the tolerant cultivar appeared to mount a relatively rapid PmKIN upregulation, whereas the sensitive one seemed to show a sluggish response and poor recovery. Three members (PmKIN22/27/35) were suggested to be potentially critical in low-temperature-response regulation. A 285-pair interaction network predicted PmKIN45 as the central hub, and functional predictions imply that PmKIN proteins may be linked to microtubules, hormone pathways, and stress signaling, possibly coordinating growth and low-temperature responses. Conclusions: This work screens candidate genes associated with low-temperature responses in P. mume, providing genetic resources for the molecular breeding of cold-resistant cultivars and expanded cultivation of P. mume and other ornamental horticultural species. Full article

Pod color is an essential economic trait of snap bean (Phaseolus vulgaris L.), and golden-yellow pods are more commercially popular than light-yellow varieties. Exploring the molecular regulatory mechanism underlying deep-yellow pod formation is essential for snap bean quality improvement. In this study, the wild-type cultivar Jinguan and the EMS-derived deep-yellow pod mutant pcm were used as materials. Integrative analyses of photosynthetic pigment quantification, genetic analysis, transcriptome sequencing, and targeted carotenoid metabolomics were performed to identify key genes and metabolic pathways governing pod color variation. Genetic analysis revealed that the deep-yellow pod trait of the pcm is stably inherited and controlled by a nuclear recessive gene. Transcriptome screening identified three differentially expressed key genes, PSY (Phvul.006G024100g), CCD4 (Phvul.002G120600g), and CYP707A (Phvul.002G122200g), whose expression patterns were strongly correlated with carotenoid accumulation. Metabolomic results demonstrated that the pcm presented reduced upstream β-carotene content, while the synthesis and accumulation of downstream chromogenic carotenoids, including zeaxanthin and capsorubin, were significantly enhanced, resulting in elevated total carotenoid content. Collectively, the differential expression of these genes remodels carotenoid metabolism and ultimately contributes to the deep-yellow pod phenotype. This study explores the regulatory mechanism of carotenoid metabolism in pod color formation, providing valuable theoretical support and genetic resources for quality breeding of snap beans. Full article

Background: Metabolic syndrome (MetS) is characterized by the clustering of central adiposity, elevated blood pressure, dysglycemia, and atherogenic dyslipidemia. Machine learning models for metabolic syndrome may show inflated performance when predictors overlap with the diagnostic criteria used to define the reference outcome. Glucose, triglycerides, and high-density lipoprotein (HDL) cholesterol are components of the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) definition of MetS; therefore, their use as predictors may introduce incorporation bias. This study aimed to evaluate whether glycated hemoglobin A1c (HbA1c) and low-density lipoprotein (LDL) cholesterol provide screening information for reference-defined MetS and to quantify the effect of predictor–outcome overlap. Methods: This retrospective cross-sectional analysis used de-identified routine-care data comprising 17,981 laboratory records, including 8982 records with MetS and 8999 without MetS. MetS was defined according to the updated NCEP ATP III criteria, with all five components available for reference classification from the same clinical encounter. The primary model used HbA1c and LDL cholesterol. For comparison, two additional models were evaluated: a criterion-component model using glucose, triglycerides, and HDL cholesterol, and a full biochemical model using all five variables. Gradient boosting was evaluated within a patient-level development and held-out internal test framework, with no patient shared between partitions. Model performance was assessed using ROC-AUC, threshold-based classification metrics, Brier score, calibration, decision curve analysis, and SHAP-based explainability. Results: The primary HbA1c–LDL cholesterol model showed moderate-to-strong discrimination for reference-defined MetS, with an ROC-AUC of 0.810, accuracy of 0.733, sensitivity of 0.824, specificity of 0.641, F1-score of 0.756, and Brier score of 0.174. The criterion-component model using glucose, triglycerides, and HDL cholesterol achieved a higher ROC-AUC of 0.935, consistent with a strong influence of predictor–outcome overlap. The full biochemical model achieved the highest ROC-AUC of 0.956 and Brier score of 0.084; however, this performance should be interpreted as an upper-bound estimate influenced by incorporation bias. SHAP analysis of the primary HbA1c–LDL cholesterol model indicated that HbA1c contributed more strongly than LDL cholesterol to the model output. Conclusions: HbA1c and LDL cholesterol provided measurable biochemical screening information for reference-defined MetS, although their discriminatory performance was lower than models including diagnostic criterion components. The substantially higher performance of models containing glucose, triglycerides, and HDL cholesterol is consistent with a substantial influence of incorporation bias when diagnostic criteria components are used as predictors. These findings should be interpreted as hypothesis-generating and limited to single-center internal validation. External, temporal, multicenter, and prospective validation is required before any clinical application can be considered. Full article

Background: Deep-learning artificial intelligence (AI) for mammographic screening moved from proof of concept to randomised evaluation in a single decade. We reviewed and meta-analysed its diagnostic accuracy and its effect on screening programmes, covering the literature published between 2015 and 2025. Methods: PubMed and Europe PMC were searched from 1 January 2015 to 31 December 2025, supplemented by ClinicalTrials.gov and by forward and backward citation searching (PRISMA 2020, PRISMA-DTA, PRISMA-S). Eligible studies evaluated a deep-learning system for cancer detection or triage in a screening population against a histopathological reference standard. Four syntheses were performed: standalone accuracy pooled on the logit-AUC scale (A), a bivariate sensitivity–specificity model (A2), cancer detection rate ratio for AI-integrated versus standard reading (B), and recall rate ratio (C). Random-effects models used restricted maximum likelihood with Knapp–Hartung intervals. Risk of bias was assessed with QUADAS-2 and QUADAS-C, and certainty with GRADE. Results: Twenty-six studies (27 reports, 2019–2025) were included. Pooled standalone AUC across 14 studies and 1,214,885 examinations was 0.890 (95% CI 0.858–0.915), with I² = 97.4% and a 95% prediction interval of 0.731–0.960. Neither publication year (p = 0.79) nor enriched versus consecutive sampling (p = 0.94) explained this dispersion in meta-regression. The bivariate model (k = 9) gave a summary sensitivity of 73.3% (64.5–80.6) at a specificity of 92.4% (86.8–95.8). Across randomised and paired prospective trials (k = 3), the pooled detection rate ratio was 1.13 (0.83–1.55), with the MASAI randomised trial alone reporting 1.29 (1.09–1.51) and a 44% reduction in screen reading. Non-randomised implementation studies (k = 5), which include a 463,094-women German programme evaluation, pooled to 1.22 (1.08–1.37) with little dispersion (I² = 19.0%). Recall changed little overall (0.95, 0.81–1.12). Certainty was very low for accuracy outcomes and moderate for the single randomised trial. Conclusions: A decade of evidence supports AI as a second reader and triage tool in organised screening, not as an autonomous replacement for the radiologist. Pooled accuracy is high on average but so dispersed that it cannot be transferred to a new programme; local validation before deployment remains necessary. The larger and more precise detection gains come from non-randomised designs, which is the pattern confounding would produce, so the randomised evidence remains the anchor. Interval-cancer and mortality endpoints are still awaited. Full article

The ITALUNG trial (Italian Lung Cancer Screening Trial) is a randomised controlled trial that evaluated the efficacy of Low-Dose Computed Tomography (LDCT) for lung cancer screening of over 3100 high-risk individuals, smokers and ex-smokers, aged 55–69. ITALUNG was coordinated by the Institute for Cancer Prevention Research in Florence and carried out across Tuscany in three main screening centres: Florence, Pisa, and Pistoia. The ITALUNG study demonstrated lower long-term lung cancer (LC) and cardiovascular mortality in the screened groups, with notable benefits observed in women. When the performance of the ITALUNG biomarker panel, i.e., loss of heterozygosity and microsatellite instability (LOH/MSI) and circulating free DNA (cfDNA), was analysed in samples of blood and sputum from asymptomatic high-risk subjects screened for lung cancer, sensitivity (SE) was 90% at baseline screening. However, the respective roles of LDCT screening and fluid biomarkers in screening for lung cancer are not established, highlighting the need for biomarkers to be able to detect tumours not only at the time of diagnosis but also when tested in pre-diagnostic samples. Therefore, we evaluated the longitudinal changes in the performance of multiple biomarkers in the ITALUNG biomarker panel and/or its single components to identify LC patients in ITALUNG. The cohort consisted of 55 LC patients whose biological specimens were collected over several years before diagnosis. Good results were found in test specimens obtained close to the time of diagnosis; out of 21 LC cases, 20 were positive in the ITALUNG biomarker panel [95% SE; 95% C.I. 0.86–1.04], 18 for LOH/MSI (86% SE; 95% C.I. 0.71–1.01) and 17 for cfDNA (85% SE; 95% C.I. 0.70–1.00). The panel was able to identify over 80% of LC cases by analysing the pre-diagnostic samples collected within 3 years prior to diagnosis: out of six LC patients, five were positive in the ITALUNG biomarker panel (83% SE; 95% C.I. 0.53–1.13). In addition, over 67% of the early and resected cancer cases were detected by the multiple-biomarker panel when biomarker testing samples were obtained within 11 years prior to diagnosis. The positivity of the ITALUNG biomarker panel might be associated with both carcinogenic field effects and early tumour presence when using pre-diagnostic samples obtained several years before diagnosis. Full article

Essential oils are difficult to disperse in hydrophilic films and may separate or volatilize during processing. We prepared chitosan quaternary ammonium salt/propolis films containing a TEMPO-oxidized cellulose nanofibril (TOCNF)/nisin-stabilized Pickering emulsion loaded with oregano essential oil. Two independent formulation variables were examined: the nisin fraction in the TOCNF/nisin particles (0–35% of particle mass) was first screened to select the emulsion stabilizer composition, whereas the selected TNO emulsion was subsequently added at 20%, 30%, 40%, or 50% (v/v) of the initial film-forming solution. At 25% nisin, encapsulation efficiency reached 88.0%, and the polydispersity index was 0.15, so this formulation was used in the films. As emulsion loading increased, the films became lighter and opacity fell from 19.75 to 4.81 A mm⁻¹. HCP/TNO30 reached a tensile strength of 45.18 MPa, but elongation at break decreased from 65.35% to 28.11% across the series. Water vapor and oxygen permeability were lowest at 40% loading. HCP/TNO50 had the highest radical-scavenging and antibacterial activities. It also reduced weight loss and quality deterioration during 12 days of green cherry tomato storage. No loading optimized every property. Mechanical strength and barrier performance peaked at 30% and 40%, respectively, while active preservation continued to improve up to 50%. Full article

Driven by global carbon neutrality targets, coal-fired power generation is undergoing substantial operational changes. Modern boilers must operate more flexibly while maintaining low emissions and reliable performance. This transition has increased the need for intelligent monitoring, operator-supervised optimization, and safety-constrained decision support. Using a reproducible search and screening procedure, this review examines the development of boiler intelligence across four interconnected technological stages. At the sensing layer, data-driven soft sensors support rapid prediction of flue gas emissions, while graph-structured spatiotemporal models are used to characterize flame and combustion states. At the modeling layer, physics-informed neural networks (PINNs) and proper orthogonal decomposition reduced-order models (POD-ROMs) are reviewed as routes for accelerating physical-field reconstruction. Surrogate models coupling computational fluid dynamics (CFD) with artificial intelligence (AI) provide another route to rapid prediction and can incorporate physical constraints. These fast field models can also serve as components of boiler digital twins for online assessment and operational guidance. They may also support early warning when abnormal conditions emerge. At the decision layer, reinforcement learning, model predictive control, and multi-objective optimization are reviewed for combustion and selective catalytic reduction (SCR) control. Because these applications are safety-critical, autonomous control must remain within actuator limits and established operating margins. Emission requirements and ammonia-slip constraints must also be satisfied. Safe deployment further requires fallback mechanisms, cybersecurity protection, and human supervision. Industrial application is still limited by data scarcity and lifecycle concept drift, while limited interpretability and simulator-to-real transfer create additional challenges. Edge latency and insufficient validation under abnormal conditions remain important barriers. Finally, industrial foundation models and large language models are discussed mainly as knowledge interfaces and operator-assistance tools rather than direct safety-critical controllers. Full article

Objectives: This study investigates the occurrence and characteristics of extended-spectrum β-lactamase (ESBL)-producing and polymyxin-resistant Escherichia coli isolated from raw milk cheese and chicken carcasses in Egypt. In addition, the study evaluates the antibacterial activity of various algal extracts from Padina pavonica and Polycladia myrica, collected from the Red Sea, Egypt, against selected virulent and multidrug-resistant E. coli isolates. Methods: A total of 200 samples, comprising raw milk cheese samples (n = 100) and chicken carcass samples (n = 100), were examined for the presence of E. coli. Antimicrobial susceptibility testing was performed for all isolates using the disk diffusion and broth microdilution techniques. Phenotypic confirmation of various resistance traits was conducted to confirm resistance patterns. PCR screening was performed for various resistance genes and virulence-associated genes. Phylogenetic grouping analysis of the E. coli isolates examined was also performed. Two algal extracts of P. pavonica and P. myrica were screened using gas chromatography–mass spectrometry and evaluated against our panel of virulent MDR foodborne E. coli isolates using the resazurin-based microtiter dilution method. Results: Among the 200 food samples, E. coli was recovered from 14% of samples and confirmed by using phoA gene detection. Of the recovered isolates, 64.3% were multidrug-resistant (MDR), with the majority originating from chicken carcasses (88.9%). Most isolates (92.9%) carried ≥3 virulence genes, with iss (100%), iutA (92.9%), and eaeA (89.3%) being the predominant determinants. Potential EPEC profiles were the most frequently detected (60.7%), followed by EHEC profiles (28.6%), whereas potential STEC/non-EHEC hybrid and ExPEC profiles were detected in 7.1% and 3.6% of isolates, respectively. The bla_CTX-M gene predominated among ESBL determinants (96.4%), while bla_SHV, bla_CMY-2, and mcr-1 were also detected. Notably, mcr-1 was identified in 82.1% of isolates, with frequent co-occurrence of mcr-1 and ESBL genes. Phylogenetic analysis revealed that most isolates belonged to group B1 (85.7%), whereas groups D and B2 accounted for 10.7% and 3.5%, respectively. In vitro antibacterial assays demonstrated that the P. pavonica methanol and ethyl acetate extracts and the P. myrica ethyl acetate extract exhibited inhibitory activity against MDR E. coli, with minimum inhibitory concentrations (MICs) ranging from 2.5 to 10 mg/mL. Overall, the P. myrica extract showed stronger antibacterial activity (MICs, 2.5–5 mg/mL) than the P. pavonica extracts (MICs, 5–10 mg/mL), highlighting their potential as natural antibacterial agents against MDR E. coli. Conclusions: These findings raise important public health concerns, as contaminated food products can act as reservoirs for the dissemination of multidrug-resistant (MDR) E coli, thereby increasing the risk of human infections ranging from mild gastrointestinal illnesses to severe, potentially life-threatening diseases. In addition, the present findings demonstrate the promising antibacterial activity of methanolic and ethyl acetate extracts of P. myrica and P. pavonica against MDR E. coli isolates, supporting their potential as sources of natural antibacterial compounds and warranting further investigation of their efficacy, mechanisms of action, and active constituents. Full article

This study proposes a lightweight domain-assisted deep learning framework for binary human presence detection using ultra-wideband (UWB) radar. The proposed methodology processes raw UWB radar signals through statistically screened, physics-grounded signal features including Fast Fourier Transform (FFT)-based frequency-domain statistics and Hilbert Transform (HT)-derived envelope statistics which are selected via a per-subject Cohen’s d screening step and stacked as auxiliary input channels alongside the raw signal for a lightweight two-dimensional convolutional neural network (2D-CNN). A cross-subject evaluation protocol (train-on-one-subject, test-on-the-other) is adopted to assess generalization across individuals rather than relying on a pooled, sample-level split. Among the candidate features, a Frequency Standard Deviation (FSTD) is shown to match or exceed the performance of every multi-feature combination tested, indicating that targeted feature selection is more consequential than input fusion for this task. To further improve deployment efficiency, post-training INT8 quantization is applied, reducing the model to approximately 23 KB while preserving classification performance for quantization-robust configurations. Hardware-in-the-loop benchmarking on the STEdgeAI platform indicates on-device inference times ranging from approximately 0.88 ms on AI-enabled STM32N6 hardware to 117–130 ms on STM32H7-class microcontrollers; these figures reflect model inference only and exclude radar acquisition and preprocessing time. Experiments are conducted on a two-subject (one male, one female) indoor dataset; the reported cross-subject results are presented as a relative comparison across feature and quantization configurations rather than as an estimate of population-level generalization. The findings nonetheless illustrate the feasibility of combining principled feature selection with quantization-aware, hardware-validated deployment on embedded artificial intelligence (AI) platforms. Full article

Pepper suffers from bacterial leaf spot and yellow leaf curl, which substantially reduce crop yield and fruit quality. Traditional manual diagnosis suffers from delayed response, subjective bias and heavy labor consumption, while existing intelligent detection pipelines lack standardized preprocessing workflows, quantitative feature screening and systematic model comparison. To fill these research gaps, we built a pepper leaf dataset with 1260 samples (healthy, bacterial spot, yellow leaf curl). Three segmentation algorithms (Lab b-channel, RGB super-green, Otsu-ACWE) were quantitatively assessed to select the optimal preprocessing scheme. We extracted 32 fused visual features (27 RGB/HSV/Lab color moments + five gray-level co-occurrence matrix (GLCM) texture metrics) and adopted a random-forest classifier to eliminate seven low-contribution redundant features, retaining 25 discriminative variables. Three representative models, namely convolutional neural network (CNN), logistic regression (LR), and genetic-algorithm-optimized back-propagation neural network (GA-BP), were constructed for parallel comparison via 20 independent repeated trials, with accuracy, precision, recall, F1-score and area under the receiver operating characteristic curve (AUC) as evaluation indicators. The results verified that Lab b-channel segmentation achieved superior background separation and intact lesion edge retention. CNN yielded the best performance, with an average test accuracy of 97.67% and an average AUC of 0.999, accompanied by minimal metric standard deviations and outstanding stability. LR exhibits low computational cost and fast training, which is promising for applications with limited computing resources. In contrast, GA-BP shows weak nonlinear fitting ability and severe prediction fluctuations, making it unsuitable for high-precision diagnosis. This study proposes a standardized experimental framework to offer theoretical guidance and algorithmic references for intelligent vegetable leaf disease identification. All experiments were conducted on a dataset collected under standardized indoor single-illumination conditions; therefore, the conclusions of this study are only applicable to such controlled scenarios. Full article

Reliable ground-based aerosol observations are essential for evaluating satellite aerosol products across heterogeneous surfaces and aerosol regimes. This study established a six-site CW193 sun photometer network in China with quality-controlled observations from January 2023 to May 2026 and assessed CW193 AOD consistency using collocated CE318 observations at Fujin and Lijiang. A consistent processing chain including Langley calibration, preprocessing, triplet consistency screening, 440 nm daily-stability screening, temporal smoothness tests, and statistical outlier removal was applied, and the sensitivity to QC threshold perturbations was examined. The quality-controlled observations were then used to evaluate the FY-3F/MERSI-III daily AOD product and three MODIS products (MCD19A2, MOD04_L2, and MYD04_L2). Across 440–1020 nm, CW193 and CE318 showed strong agreement (R = 0.979–0.999; slopes = 0.989–1.019; RMSE = 0.007–0.024), with no evident long-term drift during the available comparison periods. At Lijiang, Hebi, and Qingdao, FY-3F/MERSI-III yielded R values of 0.220, 0.664, and 0.702 and RMSE values of 0.128, 0.157, and 0.183, respectively. MODIS performance varied strongly among sites and products; MCD19A2 yielded the lowest RMSE at Dunhuang, Kashgar, and Lijiang, whereas MYD04_L2 yielded the lowest RMSE at Hebi and Qingdao. The small MOD04_L2 and MYD04_L2 matchup samples at Fujin precluded a robust ranking. All three MODIS products substantially underestimated AOD at Kashgar. Regional CW193 observations further distinguished high-AOD, low-AE conditions at Kashgar from low AOD, high-AE conditions at Lijiang. QC threshold perturbations affected data retention more than the principal aerosol and satellite evaluation statistics. These results support CW193 as a regional ground-based reference for satellite AOD evaluation while emphasizing the importance of site conditions, collocation strategy, sampling, and the empirical nature of AOD–AE optical regimes. Full article

The development of low-altitude passenger transport and emerging short-haul air services has made inter-county short-distance routes an important subject for regional aviation market research and preliminary planning. However, piloted eVTOL and low-altitude short-haul passenger services remain at an early stage, lacking continuous and stable historical operational data for supervised modelling. Moreover, unobserved county-level routes cannot be simply regarded as having no market potential. This study formulates county-level candidate route identification as a large-scale priority ranking problem under sparse proxy labels and develops a multi-relational graph ranking framework. The framework integrates county attributes, mobility, transport impedance, high-speed rail substitution conditions, and estimated air travel time variables through multiple relational graphs, and employs an MR-GAT ranking model to learn relative priorities among candidate routes using a pairwise ranking objective. Observed civil aviation route labels and flight frequencies are used as primary supervision signals, while airport catchment weak labels are introduced for auxiliary analysis. Experiments conducted over 240,350 county-level candidate ODs show that, across ten random seeds, the MR-GAT model achieves a mean internal Top-1000 recall of 16.33%, compared with 7.55% for XGBoost and 2.00% for the gravity-based heuristic. Paired bootstrap analysis further supports the higher internal top-ranked retrieval performance of MR-GAT over XGBoost. Ten-seed relational ablation shows relatively small differences among individual relation-removal variants, indicating that no single relational graph dominates the ranking performance under the current evaluation setting. Sensitivity analysis further shows that the resulting ranking is highly robust to uniform additions of up to 90 min to the estimated air travel time. Analysis of the resulting candidate set indicates that high-ranking ODs are typically associated with stronger inter-county interactions, higher ground-transport impedance, and limited direct high-speed rail connectivity. The framework therefore provides an analytical screening tool for reducing a large candidate space to a smaller set of links for subsequent market and operational feasibility assessment. Full article

Background/Objectives: Lightweight detectors are attractive for high-throughput low-dose CT lung-cancer screening, yet the training-time choices governing their accuracy are usually fixed without justification, as is the protocol used to evaluate them. We benchmarked the inter-slice input representation and the extent of bounding-box supervision for 2.5D pulmonary nodule detection under the official LUNA16 protocol. Methods: Using one YOLO11n backbone we compared a 2D central-slice baseline, thin-slab maximum-intensity projection, and adjacent-slice 2.5D input under loose and core-focused (60% of diameter) supervision. Evaluation followed the official subset0–subset9 ten-fold protocol over all 888 scans and 1186 nodules, with whole-volume inference across 227,225 axial slices, annotations_excluded.csv applied, and paired scan-level bootstrap intervals. Supervision ratio, minimum box size, negative mining, three seeds and a YOLO26n backbone were ablated. Results: The representation dominated. Adjacent-slice input reached a competition performance metric (CPM) of 0.7795 (95% CI 0.7517–0.8005) against 0.6498 for the 2D baseline (+0.1297; 10 of 10 folds; d_z = 2.96), while thin-slab projection (0.5965) was worse than a plain 2D slice. Core-focused supervision gave no benefit (−0.0209) and no ratio improved on full-diameter supervision. Re-scoring the same checkpoints over only the slices holding an annotated nodule centre reversed the supervision result (+0.0027) and shrank the representation effect fourfold (+0.0348). Conclusions: Inter-slice representation, not supervision extent, is the dominant design factor, and the slice set searched at inference decides whether either factor is measurable. Full article