Search Results (921)

Background/Objectives: Motor abnormalities are frequently observed in schizophrenia, but accessible methods for objective gait quantification remain limited. This exploratory controlled-setting study examined whether markerless smartphone-based video analysis combined with interpretable machine learning could quantify gait-related motor phenotypes in individuals with schizophrenia. Methods: Gait videos were collected from 100 individuals with schizophrenia and 35 healthy controls using a single-site, single-device, standardized recording setup. MediaPipe Pose was used to extract skeletal landmarks and derive 12 image-plane spatiotemporal and estimated two-dimensional knee-kinematic gait features. After temporal segmentation and quality control, 404 usable gait segments derived from 135 participants were analyzed as repeated segment-level observations. Decision Tree and Support Vector Machine models were applied for exploratory segment-level group-separation analysis using segment-wise 15-fold cross-validation after the full post-quality-control dataset had been balanced before fold allocation. Results: Several extracted gait features differed between groups, particularly image-plane ankle displacement, mean step displacement, displacement velocity, step characteristics, and knee-joint motion. In the Decision Tree model, image-plane ankle displacement served as the primary root node, indicating its central role in internal segment-level group separation. However, the healthy control group was substantially younger and not age-matched. In addition, the segment-level statistical comparisons did not account for within-participant clustering. Accordingly, the reported p values and confidence intervals may overstate statistical precision. Separately, segment-wise cross-validation allowed segments from the same participant to occur across folds and resampling was performed before fold partitioning. Because oversampling was performed with replacement, duplicated segment instances could also occur across training and validation folds. Consequently, the statistical findings should be interpreted as exploratory segment-level patterns rather than participant-level inference, and the machine-learning performance estimates should be regarded only as potentially optimistic apparent internal segment-level results and should not be regarded as evidence of participant-level generalization, diagnostic validity, screening accuracy, or clinical applicability. Conclusions: Markerless video-based gait analysis with interpretable machine learning may provide a feasible research-support approach for quantifying gait-related motor phenotypes in individuals with schizophrenia. These findings should not be interpreted as evidence for participant-level clinical classification, diagnostic or screening validity, clinical utility, or readiness for deployment, or as proof of cross-device or cross-environment reproducibility. Future studies require matched controls, psychiatric comparison groups, subject-wise validation, external datasets, calibrated gait measures, systematic cross-configuration reproducibility testing, and privacy-preserving data governance. Full article

Background/Objectives: Population pharmacokinetics (PopPK) has been used to aid valproic acid (VPA) dose individualization. However, this approach faces limitations owing to the complexity and high dimensionality of datasets. Machine learning (ML) can handle these challenges. However, the comparative evaluation of these ML algorithms and their practical application in optimizing VPA therapy have not yet been established. This review aims to summarize the current evidence, identify research gaps, and outline ML applications for VPA in clinical practice. Methods: PubMed, ScienceDirect, Scopus, the Association for Computing Machinery (ACM) Digital Library, and IEEE Xplore were searched from inception to October 2025. Eligible studies included original research articles using ML models to predict VPA pharmacokinetics or clinical outcomes (e.g., seizure control). Data on study design, population characteristics, features, predicted targets, ML algorithms, validation method, and model performance metrics were extracted. Results: Eleven studies were included. Most studies were retrospective, single-center designs. Ensemble tree-based models such as Random Forest, CatBoost, Gradient Boosted Regression Trees, and other ensemble methods, were consistently among the top-performing algorithms. Final models retained 3 to 19 input features, with daily dose, albumin, and body weight being the most common predictors. Only five studies performed external validation, limiting the generalizability of the models. Conclusions: Current VPA ML models demonstrated promising predictive performance. Nonetheless, most models are retrospective and single-center, with only limited external validation. Future VPA ML studies should use multicenter datasets and apply external evaluation to enhance model generalizability for clinical use. Full article

Background: Acute myeloid leukemia (AML) is a heterogeneous and aggressive hematologic malignancy, and LY-11h is a novel acylhydrazide-based histone deacetylase inhibitor with promising therapeutic potential for AML. However, its poor aqueous solubility, limited intestinal dissolution, and narrow therapeutic window hinder oral formulation development and motivate the development of dosage forms with flexible dose-design capabilities. Herein, an integrated hot-melt extrusion (HME)–fused deposition modeling (FDM) strategy was developed to convert LY-11h into printable amorphous solid dispersion (ASD) dosage forms. Methods: HPMC-AS was used as a pH-responsive carrier to enhance intestinal release while restricting premature gastric release, and HPC-EF was incorporated to improve filament processability. Single-factor and DoE studies identified critical formulation and process variables and established formulation–process–property relationships, while machine learning further modeled nonlinear interactions and guided optimization. In-line near-infrared spectroscopy combined with polarized light microscopy enabled real-time monitoring of LY-11h amorphization and melt homogenization during HME. Results: ExtraTrees and Bagging models showed promising predictive performance for key filament properties, and PAT-stage validation confirmed strong agreement with experimental values. The 15 DoE-designed ASD filaments were successfully fabricated into FDM-printed tablets with reproducible geometry. Equilibrium-solubility and in vitro dissolution studies demonstrated enhanced intestinal-pH solubility and reproducible pH-responsive release. Conclusions: Collectively, these findings establish a technological proof of concept for the manufacture of LY-11h dosage forms with adjustable formulation and geometric attributes. Further in vivo pharmacokinetic studies are required to determine whether these manufacturing capabilities translate into predictable dose–exposure relationships and individualized dose control. Full article

Introduction: Voice-based digital biomarkers have emerged as a promising approach for distinguishing individuals with neurodegenerative disorders, particularly Parkinson’s disease (PD), from healthy subjects (HS). With the increasing availability of smartphone and web-based recording tools, voice data can be collected efficiently in both clinical and remote settings. However, further validation is required before such approaches can be translated into routine clinical practice. Methods: This study used a cross-sectional analysis at the recording level, treating repeated recordings from the same participant as separate observations collected at Vito Fazzi Hospital in Lecce, Italy. Speech recordings from individuals with Parkinson’s disease (PD) and healthy controls were collected using the dedicated Talia smartphone and web application. Sustained vowel phonation (/a/) was analyzed as the primary speech task. Following data acquisition, feature extraction was performed as a crucial step in the speech analysis pipeline, as the quality and relevance of the extracted features directly influence the ability of machine learning models to discriminate between Parkinson’s disease (PD) patients and healthy controls. To capture various aspects of speech impairment associated with PD, a comprehensive set of acoustic features was extracted, including long-term features (pitch, jitter, and shimmer), nonlinear descriptors such as Recurrence Period Density Entropy (RPDE), and short-term feature based on Mel-Frequency Cepstral Coefficients (MFCCs). These features were subsequently used to develop and evaluate machine learning models for the classification of Parkinson’s disease and healthy subjects. Feature selection was performed using SHAP to identify the most informative vocal biomarkers. Model performance was assessed using five independent random train–test splits (70% training and 30% testing), supported by an internal five-fold cross-validation procedure within the training data. Multiple machine learning models were developed and evaluated, including Random Forest, Logistic Regression, Support Vector Machine, Naive Bayes, K-Nearest Neighbors, Decision Tree, Artificial Neural Network, and Gradient Boosting. Results: The evaluated models demonstrated strong recording-level classification performance. Artificial Neural Networks (ANN) and K-Nearest Neighbors (KNN) achieved the highest accuracy scores (0.9545 and 0.9494, respectively), along with superior recall (up to 0.9500), precision (up to 0.9551), and F1-score (up to 0.9525). Both models also exhibited excellent discriminative ability, with ROC-AUC values reaching 0.9882 (ANN) and 0.9893 (KNN). In contrast, Naive Bayes and Decision Tree showed comparatively lower performance across all metrics. Log-loss analysis further confirmed the robustness of ANN and KNN, which achieved the lowest values (0.2552 and 0.2510, respectively), indicating well-calibrated predictions. Overall, the findings highlight the consistency and generalizability of ANN and KNN across cross-validation splits. Conclusions: This study demonstrates that machine learning models, particularly ANN and KNN, can effectively differentiate Parkinson’s disease from healthy conditions using voice recordings. The integration of explainable AI for feature selection enhances model transparency and clinical relevance. However, the reported performance estimates were obtained from a recording-level analysis and should be interpreted as preliminary findings. Further studies involving larger cohorts and participant-level validation strategies are required to determine the generalizability and clinical applicability of these approaches. Full article

Background/Objectives: The aim of this study is to determine the extent to which early breast cancer detection behaviors in Türkiye are associated with age, education, income, and chronic disease status, and to evaluate the relationship of healthcare service utilization (contact with family physicians/general practitioners and specialists) with these behaviors. Methods: A cross-sectional secondary data analysis was conducted using nationally representative microdata from the Turkish Statistical Institute’s (TUIK) 2022 Turkiye Health Survey. The analysis included 5554 women aged 40–69. Descriptive statistics were calculated using the FERTFACTOR sample weight. Fully adjusted modified Poisson regression models were constructed for recent mammography (within the last 2 years) and monthly breast self-examination (BSE) (adjusted prevalence ratio, aPR), while Classification and Regression Tree (CART) analysis with five-fold cross-validation was applied to identify targetable risk profiles. (The maximum tree depth was set at 3, and the minimum terminal node size at 250 individuals.) Results: The rates of undergoing a recent mammogram and performing monthly BSE were 23.5% and 25.0%, respectively. Both early detection behaviors were more frequent among women who had contact with both a family physician/general practitioner and a specialist physician within the past 12 months. While contact with a specialist physician was independently associated with both behaviors, education level was positively associated with them, whereas the presence of a chronic disease was associated only with having a recent mammogram. In the CART analysis, the initial split occurred based on BSE behavior, and the prevalence of recent mammography varied approximately eightfold across the terminal nodes (cross-validated AUC = 0.681). Conclusions: In Türkiye, behaviors related to the early detection of breast cancer are associated with age, education, and contact with healthcare services. Fewer than one in four women were up to date with mammography, underscoring a substantial early detection gap. Contact with a specialist physician showed the strongest association; however, given the cross-sectional design, causality cannot be inferred and reverse causation cannot be excluded. Systematizing early detection recommendations within primary care and implementing targeted strategies for low-socioeconomic groups may help increase participation in screening. Full article

The small-data regime remains one of the most important and pressing problems in machine learning. With only a few hundred sample points in the training set, conventional ensemble methods are potentially unreliable because they have high variance, low generalization and are sensitive to individual points in the dataset, which can have a negative impact on any consumer application or safety-critical application. These problems are addressed by current regularization approaches one at a time, but there is no single approach that flexibly tunes the regularization level according to the statistics of the ensemble during training. This paper presents SelfReg-Ensemble, a variance-adaptive self-regularizing ensemble learning framework which tracks the variance at each training step of all base learners and dynamically adjusts the regularization intensity to keep the variance of the ensemble predictions at an acceptable level. The framework consists of three complementary components: (i) Variance Monitoring Module (VMM) to monitor the variance of predictions performed by the ensemble of members, (ii) Self-Regularization Controller (SRC) to adaptively map the observed variance to a regularization coefficient, using a sigmoid-bounded adaptive learning schedule, and (iii) Diversity-Preserving Aggregation Layer (DPAL) based on a weighted stacking with an entropy-regularized softmax voting mechanism, to avoid ensemble collapse. We offer rigorous theoretical analysis of the proposed framework that guarantees convergence and provides bounds on variance. These guarantees are formally established for convex, gradient-based learners; for the tree-based learners used in our experiments they serve as qualitative guidance and are supported empirically rather than formally proved. SelfReg-Ensemble is tested on 12 benchmark datasets from medical, financial and IoT domains, each with a total of fewer than 500 samples (N denotes total dataset size; effective per-fold training sizes Ntr range from 52 to 432 samples after stratified 90/10 splitting) and consistently outperforms ten state-of-the-art baselines, with on average 6.8% higher AUROC than XGBoost v2.0.3 and 5.2% higher than the actual strongest average baseline, Sub-Network Ensemble (85.6% average AUROC), 9.3% lower prediction variance, and 4.1% higher F1-score. The proposed framework is lightweight, modular and easily deployable in resource-limited consumer electronics environments. Full article

This article presents a spatial assessment of rangeland productivity, forage resources, and the potential carrying capacity of saiga (Saiga tatarica (L., 1766)) habitats within the Korgalzhyn State Nature Reserve (Kazakhstan). This research integrated field geobotanical surveys, remote sensing data, and spatial modelling approaches. During fieldwork conducted in July 2025, 44 geobotanical descriptions were completed. Long-term Sentinel-2 imagery (2015–2025), spectral vegetation indices, and ensemble machine learning algorithms (Random Forest and Gradient Boosted Decision Trees) were used to model rangeland productivity. The study area includes 16 types of rangelands. The forage resource base is primarily formed by dry-steppe communities dominated by Stipa, Festuca, and Artemisia. At the same time, halophytic vegetation containing Atriplex, Bassia, and Halocnemum provides additional forage resources under arid and saline conditions. The total annual forage reserve was estimated at 3,308,921.2 centners, and the potential annual carrying capacity was calculated at 363,033 saiga individuals across 313,868 ha. These values represent the theoretical forage-carrying capacity of the territory and should not be interpreted as actual population estimates. The results demonstrate the effectiveness of integrating remote sensing, field observations, and machine learning methods for assessing forage resources and spatial heterogeneity of rangeland ecosystems in the steppe regions of Kazakhstan. Full article

Accurate estimation of forest structural attributes over operational scales remains challenging because unmanned laser scanning (ULS) provides extensive spatial coverage but limited representation of internal stem structure, whereas terrestrial and mobile laser scanning (TLS/MLS) provide detailed stem measurements over relatively small areas. This study investigates a calibration-transfer framework in which small areas of terrestrial or fused LiDAR are used to improve diameter at breast height (DBH) estimation across much larger regions surveyed only by ULS. ULS, TLS, MLS and fused laser scanning (FLS) datasets were analysed for radiata pine and eucalyptus plantations. TreeLS-derived DBH measurements from terrestrial and fused point clouds were used as reference data to evaluate several distribution-aware and voxel-based imputation approaches for correcting regression-derived ULS estimates. Across the study sites, the best-performing imputation methods reduced stand-level mean DBH differences by as much as 95% relative to the uncorrected ULS regression estimates, resulting in substantially improved agreement with field-observed stand means while simultaneously producing DBH distributions that more closely matched the corresponding TreeLS-derived reference distributions. Voxel-based imputation performed particularly well for radiata pine and remained competitive for eucalyptus, while several distribution-based approaches achieved comparable or better performance in particular stands. These findings demonstrate the potential for transferring information from relatively small terrestrial LiDAR calibration areas to larger ULS-only acquisitions, improving stand-level DBH distribution estimates without requiring complete terrestrial coverage. Because validation was performed using stand-level field summary statistics rather than matched individual trees, the reported performance should be interpreted as demonstrating the potential of the approach under the conditions evaluated rather than universal individual-tree accuracy. Full article

The main objective of the present study was to conduct landslide susceptibility assessment and zoning analysis based on multi-scale imbalanced sampling and deep learning methods. Jiangkou Town, China, was selected as the study area. Grid cells with resolutions of 12.5 m, 25 m, and 50 m were chosen. Imbalanced sampling was applied using landslide/non-landslide ratios of 1:1, 1:2, and 1:3 to construct multi-scale modeling datasets. Susceptibility conditioning factors were screened using the frequency ratio (FR), Pearson correlation coefficient, and multicollinearity diagnostics and nine factors were obtained: slope, aspect, plane curvature, profile curvature, lithology, distance to river, distance to fault, annual rainfall, and land use. Six models—Logistic Model Tree (LMT), Kernel Logistic Regression (KLR), EfficientNet, ResNet, Transformer, and U-Net—were selected to establish 54 susceptibility evaluation models under various combinations of resolution and sampling ratios. The predictive reliability of the models was evaluated using receiver operating characteristic (ROC) curves and Kappa coefficients. Among the evaluated configurations, ResNet at a 12.5 m resolution with a 1:3 sampling ratio was retained as the preferred overall mapping configuration. It achieved a validation AUC of 0.963 and a Kappa coefficient of 0.778, together with strong susceptibility-zonation selectivity. The highest individual Kappa coefficient (0.819) was obtained by ResNet at a 25 m resolution with a 1:3 sampling ratio. Thus, the preferred configuration was identified through an integrated interpretation of the validation AUC, Kappa agreement, and susceptibility-zonation performance rather than by maximizing a single metric. The landslide susceptibility maps produced were classified into five levels and validated using the landslide distribution, landslide density and frequency ratio within each susceptibility zone. Most models showed good predictive performance in areas characterized by very high and very low susceptibility. According to the results of the comparison of the different susceptibility levels, it appears that ResNet and EfficientNet produced the most similar spatial predictions, while ResNet and Transformer presented the largest deviations. The deviations are mainly located near river valleys and areas with intense human activity. The proposed methodological framework and results can support disaster prevention, land-use planning, and regional risk management, particularly in mountainous areas with complex geological and topographic conditions. Full article

Mass cytometers can record 40–50 parameters per single cell for millions of cells in a sample. Many methods have been developed to cluster phenotypically similar cells within cytometry data, but there are fewer methods to visualize activity and interactions of pairs of proteins across these populations. We have developed a workflow for analyzing correlations associated with predefined populations. By clustering blood samples from acute myeloid leukemia (AML) patients and normal controls using an established algorithm, we obtained a minimum spanning tree of clusters of single cells. Using surface marker expression, we identified clusters on the tree that belonged to phenotypes of interest. Next, we computed correlations between pairs of proteins in each cluster. We developed a novel, coherent, probability-based statistic to test differences between vectors of correlation coefficients. By comparing all combinations of the normal controls under the statistic, we created an empirical distribution that could provide a conservative threshold of differential correlation. Using this empirically derived distribution to define significance, we compared pooled samples from AML subtypes and normal controls to detect differential correlations. Given the structure present within this cytometry dataset, we found it natural to consider correlations in this manner versus aggregating all data and computing a single correlation. Our approach has the advantage that we can localize the statistical measure to determine contributions from particular phenotypic populations. Differentially correlated pairs of proteins can be further explored as possible testable hypotheses by considering a population’s distribution of correlation coefficients or biaxially plotting protein expressions within individual cells in a given population. Full article

The increasing integration of renewable energy sources and smart grid (SG) technologies introduces significant challenges to power system stability due to the inherent variability and uncertainty of electricity generation and demand. While many machine learning and deep learning approaches have been proposed for stability prediction, various studies focus only on predictive performance and offer limited assessment of computational efficiency, statistical significance, and sustainability-related metrics. To address these gaps, this study suggests a hybrid fusion approach that combines Convolutional Neural Networks (CNNs), eXtreme Gradient Boosting (XGBoost), and Random Forest (RF) classifiers through a soft voting strategy for SG stability prediction. The CNN component automatically extracts representative features, while XGBoost and RF contribute complementary classification capabilities, reducing the need for manual feature engineering. In addition to predictive evaluation, a Green AI-oriented benchmarking framework is introduced to evaluate model performance using predictive accuracy, computational runtime, memory consumption, and estimated computational CO₂ emissions associated with model training and inference. The proposed framework is assessed on the UCI SG Stability dataset using stratified cross-validation and statistical significance testing, including Friedman and Nemenyi post hoc analyses. Experimental results demonstrate that the fusion model achieves 97.68% classification accuracy and an AUC of 0.991, exceeding several individual machine learning, deep learning, and ensemble baselines. Statistical analysis shows significant improvements over recurrent deep learning models such as LSTM and GRU. However, the differences from strong tree-based methods, including XGBoost and RF, are not statistically significant. Furthermore, the proposed model reaches a high sustainability score of 0.802, indicating a favorable balance between predictive performance and computational resource requirements. The findings show that the proposed framework is effective and computationally efficient to predict the stability of the smart grid on the UCI benchmark dataset and also serves as a transparent green AI benchmarking methodology for comparative studies in the future. The validation of the approach on real-world smart grid data under noisy, not fully complete, and heterogeneous operating conditions is another interesting research direction for future work. Full article

Urban forests play a crucial role in sustaining habitable urban environments, and tree canopy height (TCH) is a key parameter for characterizing urban forest structure and ecological functions. However, accurate TCH retrieval using spaceborne lidar remains challenging in urban built-up areas because of the high spatial heterogeneity of urban land cover. This study proposes an urban-oriented method for ICESat-2 data processing and TCH extraction and evaluates it in Peoria, Illinois, USA. By integrating spectral constraints with photon spatial distribution patterns, the method further separates non-ground photons into vegetation photons and building photons. Individual tree crown polygons are then introduced as spatial constraints for TCH extraction, helping to overcome the limitations of regular grids in representing both the laser footprint scale and fine urban landscape detail. In addition, a non-exclusive photon-to-crown assignment strategy based on energy weighting allowed each photon to contribute to multiple crowns in proportion to the footprint energy intercepted by each crown. Among the 1391 crowns associated with vegetation photons, 286 met the thresholds for both total and high-weight photon counts and were retained for accuracy assessment. Validation against canopy heights derived from airborne laser scanning (ALS) reference data shows that the proposed method achieves an R² of 0.55 and an RMSE of 2.71 m. Further analysis indicates that retrieval accuracy is primarily controlled by beam strength: strong beams yield higher accuracy (R² ≈ 0.60), whereas daytime and nighttime observations differ only slightly. Overall, after targeted processing, ICESat-2 can provide a set of individual tree canopy height samples in urban built-up areas, which may support local calibration and serve as potential labels for subsequent regional canopy height mapping. Full article

Forests play an important role in timber and fiber production, carbon storage, biodiversity conservation, and various other ecosystem services, necessitating accurate and scalable inventory methods. Recent advances in remote sensing have enabled large-scale forest monitoring; however, challenges remain in extracting reliable information across varying spatial, temporal, and environmental conditions. Deep learning has emerged as a promising tool for addressing these limitations by learning complex patterns from diverse remote sensing data sources. This review synthesizes deep learning applications in forest inventory methods across three tasks: tree counting and localization, tree species identification, and tree measurement. In total, we evaluated 122 unique primary studies (37 for tree counting and localization, 57 for species identification, and 29 for tree measurement, with one study contributing to both the counting/localization and measurement tasks) spanning terrestrial, unmanned aerial vehicle (UAV), airborne, and satellite platforms, with a primary focus on optical imagery, Light Detection and Ranging (LiDAR) data, and their fusion. Across these studies, deep learning models frequently outperformed conventional machine learning and statistical baselines, with reported gains including up to 18% improvements in biomass estimation accuracy from data fusion and individual-tree species classification accuracies exceeding 90% for select architectures. However, performance differences were influenced strongly by forest structure, species complexity, sensor capability, and validation design. Counting and localization were generally more reliable in plantations than in complex natural or urban forests, while LiDAR was particularly valuable in dense, multilayer canopies. Species-identification accuracy was highest in studies with small, distinctive species sets, whereas mixed stands with many species showed lower accuracy. Only about a third of the reviewed studies (42 of 122) were externally validated on data or sites independent of model training, and reference data for tree measurement tasks were rarely based on direct destructive sampling. External validation often revealed lower performance than within-study testing, suggesting that reported accuracies may overestimate performance in new locations or conditions. Major advances are evident in the growing use of high-resolution UAV and smartphone-based imagery for tree-level analysis, the continued value of LiDAR for structural characterization, and the increasing integration of multimodal data fusion to improve detection, classification, and measurement accuracy. Persistent challenges include the limited availability of high-quality reference data, class imbalance and inconsistent species coverage, and weak model transferability across forest types, environmental conditions, and geographic regions. Future progress will likely depend on three priorities: development of larger and more standardized labeled datasets, stronger integration of structural, spectral, and phenological information, and the design of more transferable and application-oriented deep learning frameworks. Overall, this review provides a comprehensive, quantitatively grounded overview of deep learning-driven forest inventory methods and outlines future directions for improving scalability and applicability in forest monitoring and management. Full article

The height of a water-conducting fractured zone (WCFZ) is directly related to the design of water-preserved coal mining and water-hazard risk assessment in ecologically fragile mining areas in western China. Existing empirical formulas have limited regional adaptability, and individual machine learning models may show insufficient stability under small-sample and nonlinear data conditions. To address this issue, a heterogeneous Stacking ensemble prediction framework was constructed based on measured data from the Yushen mining area. Mining thickness, working face length, mining method, burial depth, coal seam dip angle, and hard strata proportion coefficient were selected as input variables. The base layer consisted of support vector regression (SVR), classification and regression tree (CART), random forest (RF), extreme gradient boosting (XGBoost), and back-propagation neural network (BPNN), while Ridge regression was used as the meta-learner. Under the current data split, the test set R², RMSE, MAE, and MAPE of the Stacking model were 0.953, 10.99 m, 8.79 m, and 9.847%, respectively, indicating overall superiority over individual models and other ensemble configurations. The field validation results showed that the relative errors of the model for boreholes LD-1 and LD-2 in the fully mined area were 1.99% and 1.28%, respectively; however, an overestimation of 52.70% occurred for LD-3 in the coal-pillar-adjacent area. This indicates that the model is more suitable for the regional-scale screening of the maximum fractured-zone height and should not be directly used for fine-scale prediction in local boundary-affected zones. SHAP analysis showed that mining thickness, working face length, and hard strata proportion coefficient were the main influencing variables, and their response trends were generally consistent with key-strata control and the transition toward full-mining conditions. This study provides a reference for the rapid prediction of WCFZ height and preliminary evaluation of water-preserved coal mining in weakly cemented mining areas in western China. Full article

Fruit tree pruning improves orchard ventilation and light interception, thereby enhancing fruit quality and yield. Accurate measurement of pear tree branch parameters, such as branch length, growth angle, and spacing between growth points, is essential for intelligent pruning. However, pear tree canopies often exhibit severe branch crossing and adhesion, ambiguous branch-type boundaries, and difficulties in identifying target branches. To address these challenges, this study proposes a three-stage branch point cloud segmentation framework and a skeleton-based structural parameter measurement method. First, Point Transformer V3 classifies the whole-tree point cloud into annual shoots, primary branches, and the trunk. Second, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) coarsely divides the annual shoots into local adhesive clusters. Third, PlantNet performs fine instance segmentation of individual annual shoots, after which the local labels are remapped to the whole-tree coordinate space through inverse normalisation. Semantic and instance labels are then embedded into Laplacian contraction to construct the Semantic and Instance Label-Constrained Laplacian Skeleton Extraction (SILC-LSE) method. The proposed framework achieved mPrec, mRec, mF1, and mIoU values of 90.31%, 89.82%, 90.07%, and 83.76%, respectively. The mean absolute errors (MAEs) for the length estimation of annual shoots and primary branches were 0.13 m and 0.12 m, respectively. The MAE for spacing between growth points was 0.07 m, while the MAEs for growth angle estimation between different branch types were 6.70° and 7.27°, respectively. Full article