Search Results (3,451)

The present study explores the application of RNNs for the prediction and propagation of flood waves along a section of the Bârsa River, Romania, as a fast alternative to classical hydraulic models, aiming to identify new ways to alert the population. Five neural architectures were analyzed as follows: S-RNN, LSTM, GRU, Bi-LSTM, and Bi-GRU. The input data for the neural networks were derived from 2D hydraulic simulations conducted using HEC-RAS software, which provided the necessary training data for the models. It should be mentioned that the input data for the hydraulic model are synthetic hydrographs, derived from the statistical processing of recorded floods. Performance evaluation was based on standard metrics such as NSE, $R^2$ MSE, and RMSE. The results indicate that all studied networks performed well, with NSE and $R^2$ values close to 1, thus validating their capacity to reproduce complex hydrological dynamics. Overall, all models yielded satisfactory results, making them useful tools particularly the GRU and Bi-GRU architectures, which showed the most balanced behavior, delivering low errors and high stability in predicting peak discharge, water level, and flood wave volume. The GRU and Bi-GRU networks yielded the best performance, with RMSE values below 1.45, MAE under 0.3, and volume errors typically under 3%. On the other hand, LSTM architecture exhibited the most significant instability and errors, especially in estimating the flood wave volume, often having errors exceeding 9% in some sections. The study concludes by identifying several limitations, including the heavy reliance on synthetic data and its local applicability, while also proposing solutions for future analyses, such as the integration of real-world data and the expansion of the methodology to diverse river basins thus providing greater significance to RNN models. The final conclusions highlight that RNNs are powerful tools in flood risk management, contributing to the development of fast and efficient early warning systems for extreme hydrological and meteorological events. Full article

Accurate streamflow forecasting at fine temporal and spatial scales is essential to manage the diverse hydrological behaviors of individual catchments, particularly in rapidly responding mountainous regions. This study compares three forecasting models ARIMAX, LSTM, and HEC-HMS applied to the Posina River basin in northern Italy, using 13 years of hourly hydrological data. While recent literature promotes multi-basin LSTM training for generalization, we show that a well-configured single-basin LSTM, combined with a rolling forecast strategy, can achieve comparable accuracy under high-frequency, data-constrained conditions. The physically based HEC-HMS model, calibrated for continuous simulation, provides robust peak flow prediction but requires extensive parameter tuning. ARIMAX captures baseflows but underestimates sharp hydrological events. Evaluation through NSE, KGE, and MAE shows that both LSTM and HEC-HMS outperform ARIMAX, with LSTM offering a compelling balance between accuracy and ease of implementation. This study enhances our understanding of streamflow model behavior in small basins and demonstrates that LSTM networks, despite their simplified configuration, can be reliable tools for flood forecasting in localized Alpine catchments, where physical modeling is resource-intensive and regional data for multi-basin training are often unavailable. Full article

Background/Objectives: The aim of this systematic review was to compare the effectiveness of virtual reality (VR)-based training with conventional exercise programs in improving functional outcomes related to fall risk among older adults with various health conditions. Methods: The review was conducted in accordance with the PRISMA 2020 guidelines and registered in PROSPERO (registration number CRD42022345678). The databases Scopus, PubMed, Web of Science, and EBSCO were searched up to 31 March 2025. Randomized controlled trials (RCTs) were included if they involved participants aged ≥60 years, a VR intervention lasting ≥6 weeks, and a control group performing traditional exercises or receiving usual care. Methodological quality was assessed using the PEDro scale, and a narrative synthesis was performed across four outcome domains: balance, mobility, cognitive function, and fall risk. Results: Seven RCTs were included in the analysis (totaling 664 participants). VR training was found to be at least as effective as conventional exercise in improving balance (e.g., Berg Balance Scale) and mobility (e.g., Timed Up and Go), with some studies showing superior effects of VR. One RCT demonstrated that combining VR with balance exercises (MIX) yielded the greatest improvements in muscle strength and physical performance. Additionally, two studies reported cognitive benefits (e.g., MoCA) and a 42% reduction in fall incidence within six months following VR intervention. The methodological quality of the included studies was moderate to high (PEDro score 5–9/10). Conclusions: VR-based training represents a safe and engaging supplement to geriatric rehabilitation, effectively improving balance, mobility, and, in selected cases, cognitive function, while also reducing fall risk. Full article

Accurately predicting the compressive strength of high-performance concrete (HPC) is critical for ensuring structural integrity and promoting sustainable construction practices. However, HPC exhibits highly complex, nonlinear, and multi-factorial interactions among its constituents (such as cement, aggregates, admixtures, and curing conditions), which pose significant challenges to conventional predictive models. Traditional approaches often fail to adequately capture these intricate relationships, resulting in limited prediction accuracy and poor generalization. Moreover, the high dimensionality and noisy nature of HPC mix data increase the risk of model overfitting and convergence to local optima during optimization. To address these challenges, this study proposes a novel bio-inspired hybrid optimization model, AP-IVYPSO-BP, which is specifically designed to handle the nonlinear and complex nature of HPC strength prediction. The model integrates the ivy algorithm (IVYA) with particle swarm optimization (PSO) and incorporates an adaptive probability strategy based on fitness improvement to dynamically balance global exploration and local exploitation. This design effectively mitigates common issues such as premature convergence, slow convergence speed, and weak robustness in traditional metaheuristic algorithms when applied to complex engineering data. The AP-IVYPSO is employed to optimize the weights and biases of a backpropagation neural network (BPNN), thereby enhancing its predictive accuracy and robustness. The model was trained and validated on a dataset comprising 1,030 HPC mix samples. Experimental results show that AP-IVYPSO-BP significantly outperforms traditional BPNN, PSO-BP, GA-BP, and IVY-BP models across multiple evaluation metrics. Specifically, it achieved an R² of 0.9542, MAE of 3.0404, and RMSE of 3.7991 on the test set, demonstrating its high accuracy and reliability. These results confirm the potential of the proposed bio-inspired model in the prediction and optimization of concrete strength, offering practical value in civil engineering and materials design. Full article

This study develops and validates an artificial intelligence (AI)-assisted internship learning platform for automotive electronics based on the Llama 3 large language model, aiming to enhance pedagogical effectiveness within vocational training contexts. Addressing critical issues such as the persistent theory–practice gap and limited innovation capability prevalent in existing curricula, we leverage the natural language processing (NLP) capabilities of Llama 3 through fine-tuning based on transfer learning to establish a specialized knowledge base encompassing fundamental circuit principles and fault diagnosis protocols. The implementation employs the Hugging Face Transformers library with optimized hyperparameters, including a learning rate of 5 × 10⁻⁵ across five training epochs. Post-training evaluations revealed an accuracy of 89.7% on validation tasks (representing a 12.4% improvement over the baseline model), a semantic comprehension precision of 92.3% in technical question-and-answer assessments, a mathematical computation accuracy of 78.4% (highlighting this as a current limitation), and a latency of 6.3 s under peak operational workloads (indicating a system bottleneck). Although direct trials involving students were deliberately avoided, the platform’s technical feasibility was validated through multidimensional benchmarking against established models (BERT-base and GPT-2), confirming superior domain adaptability (F1 = 0.87) and enhanced error tolerance (σ² = 1.2). Notable limitations emerged in numerical reasoning tasks (Cohen’s d = 1.15 compared to human experts) and in real-time responsiveness deterioration when exceeding 50 concurrent users. The study concludes that Llama 3 demonstrates considerable promise for automotive electronics skills development. Proposed future enhancements include integrating symbolic AI modules to improve computational reliability, implementing Kubernetes-based load balancing to ensure latency below 2 s at scale, and conducting longitudinal pedagogical validation studies with trainees. This research provides a robust technical foundation for AI-driven vocational education, especially suited to mechatronics fields that require close integration between theoretical knowledge and practical troubleshooting skills. Full article

Background/Objectives: Multifactorial prevention of falls in persons with dementia has minimal or non-significant effects. Personalised prevention is recommended. We have previously shown that gait speed, basic activities of daily living (ADL), and depression (high Cornell scores) were independent predictors of falls in persons with mild and moderate cognitive impairment. This study explored person-specific risks of falls related to physical, mental, and cognitive functions and types of dementia: Alzheimer’s disease (AD), vascular dementia (VD), mixed Alzheimer’s disease/vascular dementia (MixADVD), frontotemporal dementia (FTD), and dementia with Lewy bodies (DLB). Methods: The study used data from “The Norwegian Registry of Persons Assessed for Cognitive Symptoms” (NorCog). Differences between the dementia groups and predictors of falls, gait speed, ADL, and Cornell scores were analysed. Results: Among study participants, 537/1321 (40.7%) reported a fall in the past year, with significant variations between dementia diagnoses. Fall incidence increased with age, comorbidity/polypharmacy, depression, and MAYO fluctuation score and with reduced physical activity, gait speed, and ADL. Persons with VD and MixADVD had high fall incidences and impaired gait speed and ADL. Training of physical fitness, endurance, muscular strength, coordination, and balance and optimising treatment of comorbidities and medication enhance gait speed. Improving ADL necessitates, in addition, relief of cognitive impairment and fluctuations. Relief of depression and fluctuations by psychological and pharmacological interventions is necessary to reduce the high fall risk in persons with DLB. Conclusions: The fall incidence and fall predictors varied significantly. Personalised interventions presuppose knowledge of each individual’s fall risk factors. Full article

Monitoring the height of secondary forest regrowth is essential for assessing ecosystem recovery, but current methods rely on field surveys, airborne or UAV LiDAR, and 3D reconstruction from high-resolution UAV imagery, which are often costly or limited by logistical constraints. Here, we address the challenge of scaling up canopy height monitoring by evaluating a recent deep learning model, trained on data from the Amazon and Atlantic Forests, developed to extract canopy height from RGB-NIR Planet NICFI imagery. The research questions are as follows: (i) How are canopy height estimates from the model affected by slope and orientation in natural forests, based on a large and well-balanced experimental design? (ii) How effectively does the model capture the growth trajectories of Pinus and Eucalyptus plantations over an eight-year period following planting? We find that the model closely tracks Pinus growth at the parcel scale, with predictions generally within one standard deviation of UAV-derived heights. For Eucalyptus, while growth is detected, the model consistently underestimates height, by more than 10 m in some cases, until late in the cycle when the canopy becomes less dense. In stable natural forests, the model reveals seasonal artifacts driven by topographic variables (slope × aspect × day of year), for which we propose strategies to reduce their influence. These results highlight the model’s potential as a cost-effective and scalable alternative to field-based and LiDAR methods, enabling broad-scale monitoring of forest regrowth and contributing to innovation in remote sensing for forest dynamics assessment. Full article

The Red River Jig is a traditional Métis dance practiced among Indigenous and non-Indigenous Peoples. While exercise improves physical health and fitness, the impacts of cultural dances on wholistic health are less clear. This study aimed to investigate the psychosocial (cultural and mental), social, physical function, and physical fitness benefits of a Red River Jig intervention. In partnership with Li Toneur Nimiyitoohk Métis Dance Group, Indigenous and non-Indigenous adults (N = 40, 39 ± 15 years, 32 females) completed an 8-week Red River Jig intervention. Social support, cultural identity, memory, and mental wellbeing questionnaires, seated blood pressure and heart rate, weight, pulse-wave velocity, heart rate variability, baroreceptor sensitivity, jump height, sit-and-reach flexibility, one-leg and tandem balance, and six-minute walk test were assessed pre- and post-intervention. Community, family, and friend support scores, six-minute walk distance (553.0 ± 88.7 m vs. 602.2 ± 138.6 m, p = 0.002), jump, leg power, and systolic blood pressure low-to-high-frequency ratio increased after the intervention. Ethnic identity remained the same while affirmation and belonging declined, leading to declines in overall cultural identity, as learning about Métis culture through the Red River Jig may highlight gaps in cultural knowledge. Seated systolic blood pressure (116.5 ± 7.3 mmHg vs. 112.5 ± 10.7 mmHg, p = 0.01) and lower peripheral pulse-wave velocity (10.0 ± 2.0 m·s⁻¹ vs. 9.4 ± 1.9 m·s⁻¹, p = 0.04) decreased after the intervention. Red River Jig dance training can improve social support, physical function, and physical fitness for Indigenous and non-Indigenous adults. Full article

Background/Objectives: Frailty is a prevalent geriatric syndrome associated with impaired postural control and elevated fall risk. Although conventional exercise is a core strategy for frailty management, adherence remains limited. Virtual reality (VR)-based interventions have emerged as potentially engaging alternatives, but their effects on objective postural control and task-specific confidence in frail populations remain understudied. This study aimed to evaluate the effectiveness of a supervised VR training program using the Nintendo Ring Fit Plus™ on postural control, functional mobility, and balance confidence among frail community-dwelling older adults. Methods: Fifty-one adults aged ≥65 years classified as frail or prefrail were enrolled in a four-week trial. Participants were assigned to either a VR intervention group (n = 28) or control group (n = 23). Participants were non-randomly assigned based on availability and preference. Outcome measures were collected at baseline and post-intervention. Primary outcomes included center of pressure (CoP) metrics—sway area, mean velocity, and sway path. Secondary outcomes were the Timed Up and Go (TUG), Berg Balance Scale (BBS), Activities-specific Balance Confidence (ABC), and Falls Efficacy Scale–International (FES-I). Results: After adjusting for baseline values, age, and BMI, the intervention group showed significantly greater improvements than the control group across all postural control outcomes. Notably, reductions in sway area, mean velocity, and sway path were observed under both eyes-open and eyes-closed conditions, with effect sizes ranging from moderate to very large (Cohen’s d = 0.57 to 1.61). For secondary outcomes, significant between-group differences were found in functional mobility (TUG), balance performance (BBS), and balance confidence (ABC), with moderate-to-large effect sizes (Cohen’s d = 0.53 to 0.73). However, no significant improvement was observed in fear of falling (FES-I), despite a small-to-moderate effect size. Conclusions: A supervised VR program significantly enhanced postural control, mobility, and task-specific balance confidence in frail older adults. These findings support the feasibility and efficacy of VR-based training as a scalable strategy for mitigating frailty-related mobility impairments. Full article

Background/Objectives: Diabetic retinopathy is a leading cause of vision impairment worldwide, and the development of reliable automated classification systems is crucial for early diagnosis and clinical decision-making. This study presents a comprehensive comparative evaluation of two state-of-the-art deep learning families for the task of classifying diabetic retinopathy using retinal fundus images. Methods: The models were trained and tested in both binary and multi-class settings. The experimental design involved partitioning the data into training (70%), validation (20%), and testing (10%) sets. Model performance was assessed using standard metrics, including precision, sensitivity, specificity, F1-score, and the area under the receiver operating characteristic curve. Results: In binary classification, the ResNeXt101-64x4d model and RegNetY32GT model demonstrated outstanding performance, each achieving high sensitivity and precision. For multi-class classification, ResNeXt101-32x8d exhibited strong performance in early stages, while RegNetY16GT showed better balance across all stages, particularly in advanced diabetic retinopathy cases. To enhance transparency, SHapley Additive exPlanations were employed to visualize the pixel-level contributions for each model’s predictions. Conclusions: The findings suggest that while ResNeXt models are effective in detecting early signs, RegNet models offer more consistent performance in distinguishing between multiple stages of diabetic retinopathy severity. This dual approach combining quantitative evaluation and model interpretability supports the development of more robust and clinically trustworthy decision support systems for diabetic retinopathy screening. Full article

Background and Objectives: Dual-task training (DTT) is an innovative therapeutic approach that involves the simultaneous application of two tasks, which can be motor, cognitive, or a combination of both. Children with cerebral palsy (CP) often exhibit impairments in balance, motor skills, and gait, conditions that may be amenable to improvement through DTT. The aim of this study was to determine the effectiveness of DTT in enhancing balance, walking speed, and gross motor function-related balance in children with CP. Materials and Methods: In accordance with PRISMA guidelines, a comprehensive systematic review with meta-analysis (SRMA) was conducted. Electronic databases like PubMed Medline, Scopus, Web of Science, CINAHL, and PEDro were searched up to March 2025, with no language or publication date restrictions. Only randomized controlled trials (RCTs) examining the effectiveness of DTT on balance, gross motor function, and walking speed in children with CP were included. The methodological quality and risk of bias of the included RCTs were assessed using the PEDro scale. Pooled effects were calculated using Cohen’s standardized mean difference (SMD) and its 95% confidence interval (95% CI) within random-effects models. Results: Eight RCTs, providing data from 216 children, were included. Meta-analyses suggested that DTT was more effective than conventional therapies for increasing functional (SMD = 0.65; 95% CI 0.18 to 1.13), dynamic (SMD = 0.61; 95% CI 0.15 to 1.1), and static balance (SMD = 0.46; 95% CI 0.02 to 0.9), as well as standing (SMD = 0.75; 95% CI 0.31 to 1.18; p = 0.001) and locomotion dimensions (SMD = 0.65; 95% CI 0.22 to 1.08) of the Gross Motor Function Measure (GMFM) and walking speed (SMD = 0.46; 95% CI 0.06 to 0.87). Subgroup analyses revealed that a motor–cognitive dual task is better than a motor single task for functional, dynamic, and static balance and standing and locomotion dimensions for the GMFM. Conclusions: This SRMA, including the major number of RCTs to date, suggests that DTT is effective in increasing balance, walking and gross motor function-related balance in children with CP. Full article

Inspired by the human ability to learn continuously and adapt to changing environments, researchers have proposed Online Open-World Object Detection (OLOWOD). This emerging paradigm faces the challenges of detecting known categories, discovering unknown ones, continuously learning new categories, and mitigating catastrophic forgetting. To address these challenges, we propose Category Prototype-guided Streaming Multi-Level Perturbation, PSMP, a plug-and-play method for OLOWOD. PSMP, comprising semantic-level, enhanced data-level, and enhanced feature-level perturbations jointly guided by category prototypes, operates at different representational levels to collaboratively extract latent knowledge across tasks and improve adaptability. In addition, PSMP constructs the “contrastive tension” based on the relationships among category prototypes. This mechanism inherently leverages the symmetric structure formed by class prototypes in the latent space, where prototypes of semantically similar categories tend to align symmetrically or equidistantly. By guiding perturbations along these symmetric axes, the model can achieve more balanced generalization between known and unknown categories. PSMP requires no additional annotations, is lightweight in design, and can be seamlessly integrated into existing OWOD methods. Extensive experiments show that PSMP achieves an improvement of approximately 1.5% to 3% in mAP for known categories compared to conventional online training methods while significantly increasing the Unknown Recall (UR) by around 4.6%. Full article

Anterior cruciate ligament reconstruction (ACLR) results in prolonged muscle weakness, impaired neuromuscular control, and delayed return to sport. Cross-education (CE), unilateral training of the uninjured limb, has been proposed as an adjunct therapy to promote bilateral adaptations. This scoping review evaluated the functional and neuroplastic effects of CE rehabilitation post-ACLR. Following PRISMA-ScR and JBI guidelines, PubMed, Scopus, Web of Science, and PEDro were searched up to February 2025. A bibliometric analysis was also conducted to report keyword co-occurrence and identify trends in this line of research. Of 333 screened references, 14 studies (price index: 43% and low-to-moderate risk of bias) involving 721 participants (aged 17–45 years) met inclusion criteria. CE protocols (6–12 weeks; 2–5 sessions/week) incorporating isometric, concentric, and eccentric exercises demonstrated strength gains (10–31%) and strength preservation, alongside improved limb symmetry (5–14%) and dynamic balance (7–18%). There is growing interest in neuroplasticity and corticospinal excitability, although neuroplastic changes were assessed heterogeneously across studies. Findings support CE as a feasible and low-cost strategy to complement early-stage ACLR rehabilitation, especially when direct loading of the affected limb is limited. Standardized protocols for clinical intervention and neurophysiological assessment are needed. Full article

Traditional map style transfer methods are mostly based on GAN, which are either overly artistic at the expense of conveying information, or insufficiently aesthetic by simply changing the color scheme of the map image. These methods often struggle to balance style transfer with semantic preservation and lack consistency in their transfer effects. In recent years, diffusion models have made significant progress in the field of image processing and have shown great potential in image-style transfer tasks. Inspired by these advances, this paper presents a method for transferring real-world 3D scenes to a cartoon style without the need for additional input condition guidance. The method combines pre-trained LDM with LoRA models to achieve stable and high-quality style infusion. By integrating DDIM Inversion, ControlNet, and MultiDiffusion strategies, it achieves the cartoon style transfer of real-world 3D scenes through initial noise control, detail redrawing, and global coordination. Qualitative and quantitative analyses, as well as user studies, indicate that our method effectively injects a cartoon style while preserving the semantic content of the real-world 3D scene, maintaining a high degree of consistency in style transfer. This paper offers a new perspective for map style transfer. Full article

In the study of biomechanical models, balance represents a complex problem due to the issue of indeterminate forces while standing. In order to solve this problem, it is essential to measure the ground reaction forces (GRFs) applied to each foot independently. The present work proposes a methodology for determining the independent GRF applied to each foot while standing when only one forceplate is available. For this purpose, an analytical method is proposed to determine the distribution of vertical GRFs and the position of the independent center of pressure (CoP) in each foot. Concurrently, several neural network (NN) models are trained to improve the results obtained. This hypothesis is experimentally validated by a self-developed device that allows one to simultaneously obtain the vertical GRF and CoP location of each foot at the same time that the GRF and the global CoP location are obtained from a single forceplate. The results obtained achieve a CoP position error of less than 8% and a vertical force error of 2%. The analytical hypothesis is demonstrated to offer a satisfactory level of precision, while the NN is shown to result in considerable improvement in some cases. Full article