Search Results (387)

China’s southwestern karst landscapes support remarkable cavefish diversity, especially within Sinocyclocheilus, the world’s largest cavefish genus. Using integrative taxonomic methods, we describe Sinocyclocheilus wanlanensis sp. nov., found in a subterranean river in Guizhou Province. This species lacks horn-like cranial structures; its eyes are either reduced to a dark spot or absent. It possesses a pronounced nuchal hump and a forward-protruding, duckbill-shaped head. Morphometric analysis of 28 individuals from six species shows clear separation from related taxa. Nano-CT imaging reveals distinct vertebral and cranial features. Phylogenetic analyses of mitochondrial cytb and ND4 genes place S. wanlanensis within S. angularis group as sister to S. bicornutus, with p-distances of 1.7% (cytb) and 0.7% (ND4), consistent with sister-species patterns within the genus. Sinocyclocheilus wanlanensis is differentiated from S. bicornutus by its eyeless or degenerate-eye condition and lack of bifurcated horns. It differs from S. zhenfengensis, its morphologically closest species, in having degenerate or absent eyes, shorter maxillary barbels, and pelvic fins that reach the anus. The combination of morphological and molecular evidence supports its recognition as a distinct species. Accurate documentation of such endemic and narrowly distributed taxa is important for conservation and for understanding speciation in cave habitats. Full article

Infants discriminate faces early in the first year, but research on infants’ discrimination of bodies is plagued by mixed findings. Using a familiarisation novelty preference method, we investigated 7- and 9-month-old infants’ discrimination of body postures presented in upright and inverted orientations, and with and without heads, along with relationships with gross and fine motor development. In our initial studies, 7-month-old infants discriminated upright headless postures with forward-facing and about-facing images. Eye tracking revealed that infants looked at the bodies of the upright headless postures the longest and at the heads of upright whole figures for 60–70% of the time regardless of the presence of faces, suggesting that heads detract attention from bodies. In a more stringent test, with similarly complex limb positions between test items, infants could not discriminate postures. With longer trials, the 7-month-olds demonstrated a familiarity preference for the upright whole figures, and the 9-month-olds demonstrated a novelty preference, albeit with a less robust effect. Unlike previous studies, we found that better gross motor skills were related to the 7-month-olds’ better discrimination of upright headless postures compared to inverted postures. The 9-month-old infants’ lower gross and fine motor skills were associated with a stronger preference for inverted compared to upright whole figures. This is further evidence of a configural representation of bodies in infancy, but it is constrained by an upper bias (heads in upright figures, feet in inverted), the test item similarity, and the trial duration. The measure and type of motor development reveals differential relationships with infants’ representations of bodies. Full article

Modern electronic devices such as smartphones, wearable devices, and robots typically integrate three-dimensional sensors to track the device’s movement in the 3D space. However, sensor measurements in three-dimensional vectors are highly sensitive to device orientation since a slight change in the device’s tilt or heading can change the vector values. To avoid complications, applications using these sensors often use only the magnitude of the vector, as in geomagnetic-based indoor positioning, or assume fixed device holding postures such as holding a smartphone in portrait mode only. However, using only the magnitude of the vector loses the directional information, while ad hoc posture assumptions work under controlled laboratory conditions but often fail in real-world scenarios. To resolve these problems, we propose a universal vector calibration algorithm that enables consistent three-dimensional vector measurements for the same physical activity, regardless of device orientation. The algorithm works in two stages. First, it transforms vector values in local coordinates to those in global coordinates by calibrating device tilting using pitch and roll angles computed from the initial vector values. Second, it additionally transforms vector values from the global coordinate to a reference coordinate when the target coordinate is different from the global coordinate by correcting yaw rotation to align with application-specific reference coordinate systems. We evaluated our algorithm on geomagnetic field-based indoor positioning and bidirectional step detection. For indoor positioning, our vector calibration achieved an 83.6% reduction in mismatches between sampled magnetic vectors and magnetic field map vectors and reduced the LSTM-based positioning error from 31.14 m to 0.66 m. For bidirectional step detection, the proposed algorithm with vector calibration improved step detection accuracy from 67.63% to 99.25% and forward/backward classification from 65.54% to 100% across various device orientations. Full article

To address the challenges of low recognition accuracy and substantial model complexity in crop disease identification models operating in complex field environments, this study proposed a novel hybrid model named ConvTransNet-S, which integrates Convolutional Neural Networks (CNNs) and transformers for crop disease identification tasks. Unlike existing hybrid approaches, ConvTransNet-S uniquely introduces three key innovations: First, a Local Perception Unit (LPU) and Lightweight Multi-Head Self-Attention (LMHSA) modules were introduced to synergistically enhance the extraction of fine-grained plant disease details and model global dependency relationships, respectively. Second, an Inverted Residual Feed-Forward Network (IRFFN) was employed to optimize the feature propagation path, thereby enhancing the model’s robustness against interferences such as lighting variations and leaf occlusions. This novel combination of a LPU, LMHSA, and an IRFFN achieves a dynamic equilibrium between local texture perception and global context modeling—effectively resolving the trade-offs inherent in standalone CNNs or transformers. Finally, through a phased architecture design, efficient fusion of multi-scale disease features is achieved, which enhances feature discriminability while reducing model complexity. The experimental results indicated that ConvTransNet-S achieved a recognition accuracy of 98.85% on the PlantVillage public dataset. This model operates with only 25.14 million parameters, a computational load of 3.762 GFLOPs, and an inference time of 7.56 ms. Testing on a self-built in-field complex scene dataset comprising 10,441 images revealed that ConvTransNet-S achieved an accuracy of 88.53%, which represents improvements of 14.22%, 2.75%, and 0.34% over EfficientNetV2, Vision Transformer, and Swin Transformer, respectively. Furthermore, the ConvTransNet-S model achieved up to 14.22% higher disease recognition accuracy under complex background conditions while reducing the parameter count by 46.8%. This confirms that its unique multi-scale feature mechanism can effectively distinguish disease from background features, providing a novel technical approach for disease diagnosis in complex agricultural scenarios and demonstrating significant application value for intelligent agricultural management. Full article

Background/Objectives: A randomized controlled trial (RCT) was designed to test the emerging role of respiratory mechanics as part of physiotherapy in patients with non-specific chronic neck pain (NSCNP). Methods: Ninety patients with NSCNP and symptom duration >3 months were randomly allocated to three intervention groups of equal size, receiving either cervical spine (according to the Mulligan Concept) and diaphragm manual therapy plus breathing reeducation exercises (experimental group—EG1), cervical spine manual therapy plus sham diaphragmatic manual techniques (EG2), or conventional physiotherapy (control group—CG). The treatment period lasted one month (10 sessions) for all groups. The effect on the cervical spine range of motion (CS-ROM) and on the craniovertebral angle (CVA) was examined. Outcomes were collected before treatment (0/12), after treatment (1/12), and three months after the end of treatment (4/12). The main analysis comprised a two-way mixed ANOVA with a repeated measures factor (time) and a between-groups factor (group). Post hoc tests assessed the source of significant interactions detected. The significance level was set at p = 0.05. Results: No significant between-group baseline differences were identified. Increases in CS-ROM and in CVA were registered mainly post-treatment, with improvements maintained at follow-up for CS-ROM. EG1 significantly improved over CG in all movement directions except for flexion and over EG2 for extension only, at 1/12 and 4/12. All groups improved by the same amount for CVA. Conclusions: EG1, which included diaphragm manual therapy and breathing re-education exercises, registered the largest overall improvement over CG (except for flexion and CVA), and for extension over EG2. The interaction between respiratory mechanics and neck mobility may provide new therapeutic and assessment insights of patients with NSCNP. Full article

In recent years, deep learning (DL) has been demonstrated remarkable capabilities in hyperspectral unmixing (HU) due to its powerful feature representation ability. Convolutional neural networks (CNNs) are effective in capturing local spatial information, but limited in modeling long-range dependencies. In contrast, transformer architectures extract global contextual features via multi-head self-attention (MHSA) mechanisms. However, most existing transformer-based HU methods focus only on spatial or spectral modeling at a single scale, lacking a unified mechanism to jointly explore spatial and channel-wise dependencies. This limitation is particularly critical for multiscale contextual representation in complex scenes. To address these issues, this article proposes a novel Spatial-Channel Multiscale Transformer Network (SCMT-Net) for HU. Specifically, a compact feature projection (CFP) module is first used to extract shallow discriminative features. Then, a spatial multiscale transformer (SMT) and a channel multiscale transformer (CMT) are sequentially applied to model contextual relations across spatial dimensions and long-range dependencies among spectral channels. In addition, a multiscale multi-head self-attention (MMSA) module is designed to extract rich multiscale global contextual and channel information, enabling a balance between accuracy and efficiency. An efficient feed-forward network (E-FFN) is further introduced to enhance inter-channel information flow and fusion. Experiments conducted on three real hyperspectral datasets (Samson, Jasper and Apex) and one synthetic dataset showed that SCMT-Net consistently outperformed existing approaches in both abundance estimation and endmember extraction, demonstrating superior accuracy and robustness. Full article

Accurate estimation of rice leaf nitrogen content (LNC) is essential for optimizing nitrogen management in precision agriculture. However, challenges such as spectral saturation and canopy structural variations across different growth stages complicate this task. This study proposes a robust framework for LNC estimation that integrates both spectral and texture features extracted from UAV-based multispectral imagery through the development of novel Spectral–Texture Fusion Indices (STFIs). Field data were collected under nitrogen gradient treatments across three critical growth stages: heading, early filling, and late filling. A total of 18 vegetation indices (VIs), 40 texture features (TFs), and 27 STFIs were derived from UAV images. To optimize the feature set, a two-stage feature selection strategy was employed, combining Pearson correlation analysis with model-specific embedded selection methods: Recursive Feature Elimination with Cross-Validation (RFECV) for Random Forest (RF) and Extreme Gradient Boosting (XGBoost), and Sequential Forward Selection (SFS) for Support Vector Regression (SVR) and Deep Neural Networks (DNNs). The models—RFECV-RF, RFECV-XGBoost, SFS-SVR, and SFS-DNN—were evaluated using four feature configurations. The SFS-DNN model with STFIs achieved the highest prediction accuracy (R² = 0.874, RMSE = 2.621 mg/g). SHAP analysis revealed the significant contribution of STFIs to model predictions, underscoring the effectiveness of integrating spectral and texture information. The proposed STFI-based framework demonstrates strong generalization across phenological stages and offers a scalable, interpretable approach for UAV-based nitrogen monitoring in rice production systems. Full article

Objectives: New smartphone-based methods for measuring cervical spine range of motion (CS-ROM) and posture are emerging. The purpose of this study was to assess the reliability and validity of three such methods in patients with non-specific chronic neck pain (NSCNP). Methods: The within-day test–retest reliability of CS-ROM and forward head posture (craniovertebral angle-CVA) was examined in 45 patients with NSCNP. CS-ROM was simultaneously measured with an accelerometer sensor (KFORCE Sens^®) and a mobile phone device (iHandy and Compass apps), testing the accuracy of each and the parallel-forms reliability between the two methods. For construct validity, correlations of CS-ROM with demographics, lifestyle, and other cervical and thoracic spine biomechanically based measures were examined in 90 patients with NSCNP. Male–female differences were also explored. Results: Both methods were reliable, with measurements concurring between the two devices in all six movement directions (intraclass correlation coefficient/ICC = 0.90–0.99, standard error of the measurement/SEM = 0.54–3.09°). Male–female differences were only noted for two CS-ROM measures and CVA. Significant associations were documented: (a) between the six CS-ROM measures (R = 0.22–0.54, p < 0.05), (b) participants’ age with five out of six CS-ROM measures (R = 0.23–0.40, p < 0.05) and CVA (R = 0.21, p < 0.05), (c) CVA with two out of six CS-ROM measures (extension R = 0.29, p = 0.005 and left-side flexion R = 0.21, p < 0.05), body mass (R = −0.39, p < 0.001), body mass index (R = −0.52, p < 0.001), and chest wall expansion (R = 0.24–0.29, p < 0.05). Significantly lower forward head posture was noted in subjects with a high level of physical activity relative to those with a low level of physical activity. Conclusions: The reliability of both CS-ROM methods was excellent. Reductions in CS-ROM and increases in CVA were age-dependent in NSCNP. The significant relationship identified between CVA and CWE possibly signifies interconnections between NSCNP and the biomechanical aspect of dysfunctional breathing. Full article

Minimizing energy consumption remains a critical challenge in wireless sensor networks (WSNs) because of their reliance on nonrechargeable batteries. Clustering-based hierarchical communication has been widely adopted to address this issue by improving residual energy and balancing the network load. In this architecture, cluster heads (CHs) are responsible for data collection, aggregation, and forwarding, making their optimal selection essential for prolonging network lifetime. The effectiveness of CH selection is highly dependent on the choice of metaheuristic optimization method and the design of the fitness function. Although numerous studies have applied metaheuristic algorithms with suitably designed fitness functions to tackle the CH selection problem, many existing approaches fail to fully capture both the spatial distribution of nodes and dynamic energy conditions. To address these limitations, we propose the binary secretary bird optimization clustering (BSBOC) method. BSBOC introduces a binary variant of the secretary bird optimization algorithm (SBOA) to handle the discrete nature of CH selection. Additionally, it defines a novel multiobjective fitness function that, for the first time, considers the Voronoi diagram of CHs as an optimization objective, besides other well-known objectives. BSBOC was thoroughly assessed via comprehensive simulation experiments, benchmarked against two advanced methods (MOBGWO and WAOA), under both homogeneous and heterogeneous network models across two deployment scenarios. Findings from these simulations demonstrated that BSBOC notably decreased energy usage and prolonged network lifetime, highlighting its effectiveness as a reliable method for energy-aware clustering in WSNs. Full article

Efficient and timely data collection in Underwater Acoustic Sensor Networks (UASNs) for Internet of Underwater Things (IoUT) applications remains a significant challenge due to the inherent limitations of the underwater environment. This paper presents a Value of Information (VoI)-based trajectory planning framework for a single Autonomous Underwater Vehicle (AUV) operating in coordination with an Unmanned Surface Vehicle (USV) to collect data from multiple Cluster Heads (CHs) deployed across an uneven seafloor. The proposed approach employs a VoI model that captures both the importance and timeliness of sensed data, guiding the AUV to collect and deliver critical information before its value significantly degrades. A forward Dynamic Programming (DP) algorithm is used to jointly optimize the AUV’s trajectory and the USV’s start and end positions, with the objective of maximizing the total residual VoI upon mission completion. The trajectory design incorporates the AUV’s kinematic constraints into travel time estimation, enabling accurate VoI evaluation throughout the mission. Simulation results show that the proposed strategy consistently outperforms conventional baselines in terms of residual VoI and overall system efficiency. These findings highlight the advantages of VoI-aware planning and AUV–USV collaboration for effective data collection in challenging underwater environments. Full article

Graph attention networks are pivotal for modeling non-Euclidean data, yet they face dual challenges: training oscillations induced by projection-based high-dimensional constraints and gradient anomalies due to poor adaptation to heterophilic structure. To address these issues, we propose LDC-GAT (Lyapunov-Stable Graph Attention Network with Dynamic Filtering and Constraint-Aware Optimization), which jointly optimizes both forward and backward propagation processes. In the forward path, we introduce Dynamic Residual Graph Filtering, which integrates a tunable self-loop coefficient to balance neighborhood aggregation and self-feature retention. This filtering mechanism, constrained by a lower bound on Dirichlet energy, improves multi-head attention via multi-scale fusion and mitigates overfitting. In the backward path, we design the Fro-FWNAdam, a gradient descent algorithm guided by a learning-rate-aware perceptron. An explicit Frobenius norm bound on weights is derived from Lyapunov theory to form the basis of the perceptron. This stability-aware optimizer is embedded within a Frank–Wolfe framework with Nesterov acceleration, yielding a projection-free constrained optimization strategy that stabilizes training dynamics. Experiments on six benchmark datasets show that LDC-GAT outperforms GAT by 10.54% in classification accuracy, which demonstrates strong robustness on heterophilic graphs. Full article

Advances in information technology and the widespread adoption of smartphones have improved human convenience and quality of life by facilitating extensive information sharing. However, the increasing frequency and duration of smartphone use is linked to a high risk of musculoskeletal disorders, particularly manifesting as changes in posture and gait. These alterations can lead to various physical issues, including spinal deformities, reduced gait stability, and increased muscle fatigue. Furthermore, excessive smartphone use can negatively affect mental health, contributing to depression, anxiety, and cognitive impairment. This narrative review primarily aims to systematically examine the effects of smartphone-related posture and gait alterations on physical function and identify associated problems. This study systematically summarized individual studies published between 2009, when smartphones first became widespread, and 2024 that investigated the effects of smartphone-induced posture and gait alterations. Through identifying issues related to these alterations, we aim to propose preventive strategies to avoid further complications. Full article

The present study aimed to evaluate the effects of self-loading training on the overall stress response during pre-slaughter transportation in slaughter horses. Thirty-two slaughter horses were divided into two groups: the control group (CG) and the Trained Group (TG). For six weeks, the TG horses were trained to self-load using a method based on target training and shaping. Animals from both groups were transported to the same slaughterhouse in small groups on different days using the same truck along the same route. The baseline and post-transportation values of the eye temperature and fecal cortisol metabolites were determined for all the animals. All the horses were video-recorded while being transported and unloaded. The horses’ behavior was analyzed using a focal animal sampling method. During transportation, the presence of head shaking, licking, and chewing was significantly higher in the CG than the TG. During unloading, walking forward tended to be more frequent in the TG. The fecal cortisol metabolites and eye temperature were higher after transportation than at the baseline, but no differences between the groups were found. Overall, the results confirm that transportation for less than one hour caused detectable stress in slaughter horses. Based on these results, self-loading training appears to be somewhat useful to mitigate the overall transport stress. Full article

Background: The purpose of the proposed study is to investigate the efficacy of UNet in predicting Deoxyribonucleic Acid methylation patterns in a cervical cancer cell line. The application of deep learning to analyse the factors affecting methylation in the context of cervical cancer has not yet been fully explored. Methods: A comprehensive performance evaluation has been conducted based on multiple window sizes of DNA sequences. For this purpose, three different parameter-analysis techniques, namely, autoencoders, Generative Adversarial Networks, and Multi-Head Attention Networks, were used. This work presents a novel framework for methylation prediction in promoter regions of various genes. Results and Conclusions: Experimental results have proved that attention networks in association with UNet achieved a significant accuracy level of 91.01% along with a sensitivity of 89.65%, specificity of around 92.35%, and an area under curve of 0.910 on ENCODE database. The proposed model outperformed three state-of-the-art models: Convolutional Neural Network, Transfer Learning, and Feed Forward Neural Network with K-Nearest Neighbour. Moreover, validation of the model in five gene promoters achieved an accuracy of 81.60% with an area under curve score of 0.814, a p-value of $3.62\times {10}^{-19}$, and Cohen’s Kappa value of 0.631. This novel approach has led to a better understanding of epigenetic variables and their implications in cervical cancer, offering potential insights into therapeutic strategies. Full article

Background/Objectives: Accurate postural assessment is essential for managing musculoskeletal disorders; however, routine screening is often limited by radiation exposure, cost, and accessibility constraints of radiography. Recent advances in artificial intelligence (AI) have enabled automated, marker-free analysis using two-dimensional photographs. This study evaluated the validity and reliability of MORA Vu, an AI-based posture estimation software, against radiographic parameters. Methods: A prospective pilot study was conducted with 72 participants, divided equally into the cervical and lower-limb alignment groups. Forward head posture (FHP) and digital hip–knee–ankle (DHKA) angles were measured using MORA Vu and compared with corresponding radiographic parameters. Three healthcare professionals independently conducted the AI-based assessments. Correlations were analyzed, and interrater reliability was assessed using the intraclass correlation coefficient (ICC). Results: FHP showed the strongest correlation with the craniovertebral angle (r = −0.712) and C2–7 sagittal vertical axis (r = 0.704). The DHKA angle strongly correlated with the radiographic hip–knee–ankle angle (r = 0.754). Interrater reliability demonstrated high agreement (ICC: 0.84 FHP, 0.90 DHKA). Conclusions: MORA Vu demonstrated strong validity and high reliability, supporting its potential as a noninvasive screening tool for postural assessment. Given its accessibility and radiation-free nature, it may serve as a viable alternative for routine postural evaluation. Full article