Search Results (2,471)

Background: Correlation of genes within tissues has attracted much attention. In contrast, genes that are INDependent In Expression (INDIE) remain poorly understood, even though they may represent tissue admixtures, reflect new regulatory mechanisms, either transcriptional or post-transcriptional, and contribute to biomarkers or machine learning algorithms. We hypothesised that INDIE genes can be found, may remain uncorrelated across tissues, and replicate within tissues in external datasets. Methods: Biweight midcorrelation was calculated for each gene against all other genes with sufficiently high expression in the given tissue from the GTEx dataset v8, along with the means of absolute values of obtained correlation coefficients. The threshold for gene designation as INDIE was both absolute (r) and relative (Z-score), while the threshold for external validation in the whole blood (four datasets) and the ileum (two datasets) was relative. Results: Only one gene, RPL13P12, was INDIE in all the analysed GTEx tissues, but it did not replicate in the external datasets. In contrast, HIST1H2AD and TMEM176B were not only INDIE in GTEx whole blood but also replicated in all four external datasets, despite their heterogeneity. Moreover, ACAT2 replicated in both external ileal datasets. The haemoglobin gene HBB belonged to most widespread INDIE genes in various GTEx tissues and was validated in an external ileal dataset, pointing towards the importance of tissue heterogeneity in bulk samples. Conclusions: A set of genes exhibiting independent expression patterns across various tissues of GTEx was described. Results for each tissue are made available. Even though many findings can be explained by tissue heterogeneity, some results point towards interesting mechanisms of gene expression regulation. Full article

Background: To propose an automatic liver and hepatic vessel segmentation solution based on a stacking model and decision fusion. This model combines the decisions of multiple models to achieve increased accuracy. It exhibits improved robustness due to the reduction of individual errors. Flexibility is also a key asset, with combination methods such as majority voting or weighted averaging. The model enables managing the uncertainty associated with individual decisions to obtain a more reliable final decision. The combination of decisions improves the overall accuracy of the system. Methods: This research introduces a new deep learning-based architecture for automatically segmenting hepatic vessels and tumors from CT scans, utilizing stacking, decision fusion, and deep transfer learning to achieve high-accuracy and rapid segmentation. This study employed two distinct datasets: the external “Medical Segmentation Decathlon (MSD) task 08” dataset and an internal dataset procured from Ibn Sina University Hospital encompassing a cohort of 112 patients with chronic liver disease who underwent contrast-enhanced abdominal CT scans. Results: The proposed segmentation model reached a DSC of 83.21 and an IoU of 72.76 for hepatic vasculature and tumor segmentation, thereby exceeding the performance benchmarks established by the majority of antecedent studies. Conclusions: This study introduces an automated method for liver vessels and liver tumor segmentation, combining precision and stability to bridge the clinical gap. Furthermore, decision fusion-based stacking models have a significant impact on clinical applications by enhancing diagnostic accuracy, enabling personalized care through the integration of genetic, environmental, and clinical data, optimizing clinical trials, and facilitating the development of personalized medicines and therapies. Full article

This work presents a proof of concept including simulation and experimental validations of acoustic gas sensor prototypes for trace CO₂ detection up to 1 ppm. For the detection of lower gas concentrations especially, the dependency of acoustic resonances on the molecular weights and, consequently, the speed of sound of the gas mixture, is exploited. We explored two resonator types: a cylindrical acoustic resonator and a Helmholtz resonator intrinsic to the MEMS microphone’s geometry. Both systems utilized mass flow controllers (MFCs) for precise gas mixing and were also modeled in COMSOL Multiphysics 6.2 to simulate resonance shifts based on thermodynamic properties of binary gas mixtures, in this case, N₂-CO₂. We performed experimental tracking using Zurich Instruments MFIA, with high-resolution frequency shifts observed in µHz and mHz ranges in both setups. A compact and geometry-independent nature of MEMS-based Helmholtz tracking showed clear potential for scalable sensor designs. Multiple experimental trials confirmed the reproducibility and stability of both configurations, thus providing a robust basis for statistical validation and system reliability assessment. The good simulation experiment agreement, especially in frequency shift trends and gas density, supports the method’s viability for scalable environmental and industrial gas sensing applications. This resonance tracking system offers high sensitivity and flexibility, allowing selective detection of low CO₂ concentrations down to 1 ppm. By further exploiting both external and intrinsic acoustic resonances, the system enables highly sensitive, multi-modal sensing with minimal hardware modifications. At microscopic scales, gas detection is influenced by ambient factors like temperature and humidity, which are monitored here in a laboratory setting via NDIR sensors. A key challenge is that different gas mixtures with similar sound speeds can cause indistinguishable frequency shifts. To address this, machine learning-based multivariate gas analysis can be employed. This would, in addition to the acoustic properties of the gases as one of the variables, also consider other gas-specific variables such as absorption, molecular properties, and spectroscopic signatures, reducing cross-sensitivity and improving selectivity. This multivariate sensing approach holds potential for future application and validation with more critical gas species. Full article

Risk prediction models are increasingly used in the management of coronary artery disease (CAD), with applications ranging from diagnostic stratification to prognostic assessment and therapeutic guidance. In the context of CAD and percutaneous coronary intervention, clinical decision-making often relies on risk scores to estimate the likelihood of ischemic and bleeding events and to tailor antithrombotic strategies accordingly. Traditional scores are derived from clinical, anatomical, procedural, and laboratory variables, and their performance is evaluated based on discrimination and calibration metrics. While many established models are simple, interpretable, and externally validated, their predictive ability is often moderate and may be limited by outdated derivation cohorts, overfitting, or lack of generalizability. Recent advances have introduced artificial intelligence and machine learning models that can process large, high-dimensional datasets and identify patterns not apparent through conventional methods, with the aim to incorporate complex data; however, they are not exempt from limitations and struggle with integration into clinical practice. Notably, ethical issues, such as equity in model application, over-stratification, and real-world implementation, are of critical importance. The ideal predictive model should be accurate, generalizable, and clinically actionable. This review aims at providing an overview of the main predictive models used in the field of CAD and to discuss methodological challenges, with a focus on strengths, limitations and areas of applicability of predictive models. Full article

The significance of children’s mathematical competence during the early years is well established; however, the methods for developing such competencies remain less understood. Specifically, there is a need to identify what constitutes high-quality educational environments and effective instruction. Both the study and promotion of high-quality educational environments and teaching, through coaching and other professional development initiatives, necessitate the use of observational instruments that are reliable, efficient, and valid, including content, internal, external, and consequential validity. Moreover, domain-specific measures are essential, as general quality measures often fail to adequately assess curriculum content, scope, or sequence, and they do not reliably predict improvements in children’s learning outcomes. This study employed innovative analytical techniques to evaluate the scoring and interpretation of an existing domain-specific observational measure: the Classroom Observation of Early Mathematics Environment and Teaching (COEMET). We applied non-linear modeling approaches (i.e., Random Forest [RF] and Generalized Additive Models [GAMs]) to investigate and provide a comprehensive overview of the relationships between COEMET’s measures—at both the scale and item levels—of teachers’ practices and children’s mathematical competencies. The study first employed the RF machine learning method to identify the most important COEMET items for prediction, followed by the use of GAMs to depict the non-linear relationships between COEMET predictors and the outcome variable. The analysis revealed that certain teaching practices, as indicated by the COEMET items, exhibited non-linear and even non-monotonic associations with children’s mathematical competencies. Full article

Purpose: Accurate survival prediction is essential for optimizing the treatment planning in patients with castration-resistant prostate cancer (CRPC). However, the traditional statistical models often underperform due to limited variable inclusion and an inability to account for complex, multidimensional data interactions. Methods: We retrospectively collected 46 clinical, laboratory, and pathological variables from 801 patients with CRPC, covering the disease course from the initial disease diagnosis to CRPC progression. Multiple machine learning (ML) models, including random survival forests (RSFs), XGBoost, LightGBM, and logistic regression, were developed to predict cancer-specific mortality (CSM), overall mortality (OM), and 2- and 3-year survival status. The dataset was split into training and test cohorts (80:20), with 10-fold cross-validation. The performance was assessed using the C-index for regression models and the AUC, accuracy, precision, recall, and F1-score for classification models. Model interpretability was assessed using SHapley Additive exPlanations (SHAP). Results: Over a median follow-up of 24 months, 70.6% of patients experienced CSM. RSFs achieved the highest C-index in the test set for both CSM (0.772) and OM (0.771). For classification tasks, RSFs demonstrated a superior performance in predicting 2-year survival, while XGBoost yielded the highest F1-score for 3-year survival. The SHAP analysis identified time to first-line CRPC treatment and hemoglobin and alkaline phosphatase levels as key predictors of survival outcomes. Conclusion: The RSF and XGBoost ML models demonstrated a superior performance over that of traditional statistical methods in predicting survival in CRPC. These models offer accurate and interpretable prognostic tools that may inform personalized treatment strategies. External validation and the integration of emerging therapies are warranted for broader clinical applicability. Full article

Parkinson’s disease (PD) is one of the most rapidly growing neurological disorders globally. The molecular relationship between common respiratory infections (RIs) and idiopathic Parkinson’s disease (iPD) remains a controversial issue. Multiple studies have linked acute respiratory infections to PD, but the molecular mechanism behind this connection is not significantly defined. Therefore, the aim of our study was to investigate potential molecular interactions between RIs and PD. We retrieved eight publicly available RNA-seq datasets from the NCBI Gene Expression Omnibus (NCBI GEO) and performed extensive bioinformatics analysis, including differential gene expression (DGE) analysis, the identification of overlapped differentially expressed genes (DEGs), weighted gene co-expression network analysis (WGCNA), pathway and functional enrichment analysis, the construction of protein–protein networks, and the identification of hub genes. Additionally, we applied a machine learning method, a Random Forest model (RF), to external RIs datasets to identify the most important genes. We found that ribosomal subunits, mitochondrial complex proteins, proteasome subunits, and proteins encoding ubiquitin are simultaneously downregulated and co-expressed in RIs and PD. Dysregulation of these proteins may disturb multiple pathways, such as those responsible for ribosome biogenesis, protein synthesis, autophagy, and apoptosis; the ubiquitin–proteasome system (UPS); and the mitochondrial respiratory chain. These processes have been implicated in PD’s pathology, namely in the aggregation of α-synuclein, mitochondrial dysfunction, and the death of dopaminergic neuron cells. Our findings suggest that there are significant similarities in transcriptional responses and dysfunctional molecular mechanisms between RIs, PD, and aging. RIs may modulate PD-relevant pathways in an age- or immune-dependent manner; longitudinal studies are needed to examine the RIs risk factor. Therefore, future studies should experimentally investigate the influence of age, vaccination status, infection type, and severity to clarify the role of RIs in PD’s pathogenesis. Full article

This study analyzed subjective theories on school violence and convivencia escolar expressed in the public discourse of the Chilean Ministry of Education in 2022. This research focused on the return to in-person learning, a time when concerns about violence in schools increased and public policies aimed at addressing it were launched. Inductive content analysis and grounded theory techniques were used to examine 66 tweets issued by official ministry accounts during 2022. The analysis identified three interpretative sets. The first suggests that although violence has external structural causes, it must be eradicated from schools. The second links convivencia escolar with well-being and socioemotional skills, but without an explicit association with violence. The third locates the origin of psychological distress in external factors but assigns its management to the school system. A predominance of expert knowledge existed in the promoted solutions. These findings are discussed based on the idea that the Ministry of Education’s discourse on Twitter not only informs but also seeks to shape educational common sense and validate public policies. This raises questions about its impact on the interpretive autonomy of school communities. Full article

Accurate shade matching is essential in restorative and prosthetic dentistry yet remains difficult due to subjectivity in visual assessments. We develop and evaluate a deep learning approach for the simultaneous segmentation of natural teeth and shade guides in intraoral photographs using four fine-tuned variants of Segment Anything Model 2 (SAM2: tiny, small, base plus, and large) and a UNet baseline trained under the same protocol. The spatial performance was assessed using the Dice Similarity Coefficient (DSC), the Intersection over the Union (IoU), and the 95th-percentile Hausdorff distance normalized by the ground-truth equivalent diameter (HD95). The color consistency within masks was quantified by the coefficient of variation (CV) of the CIELAB components (L*, a*, b*). The perceptual color difference was measured using CIEDE2000 (ΔE00). On a held-out test set, all SAM2 variants achieved a high overlap accuracy; SAM2-large performed best (DSC: 0.987 ± 0.006; IoU: 0.975 ± 0.012; HD95: 1.25 ± 1.80%), followed by SAM2-small (0.987 ± 0.008; 0.974 ± 0.014; 2.96 ± 11.03%), SAM2-base plus (0.985 ± 0.011; 0.971 ± 0.021; 1.71 ± 3.28%), and SAM2-tiny (0.979 ± 0.015; 0.959 ± 0.028; 6.16 ± 11.17%). UNet reached a DSC = 0.972 ± 0.020, an IoU = 0.947 ± 0.035, and an HD95 = 6.54 ± 16.35%. The CV distributions for all of the prediction models closely matched the ground truth (e.g., GT L*: 0.164 ± 0.040; UNet: 0.144 ± 0.028; SAM2-small: 0.164 ± 0.038; SAM2-base plus: 0.162 ± 0.039). The full-mask ΔE00 was low across models, with the summary statistics reported as the median (mean ± SD): UNet: 0.325 (0.487 ± 0.364); SAM2-tiny: 0.162 (0.410 ± 0.665); SAM2-small: 0.078 (0.126 ± 0.166); SAM2-base plus: 0.072 (0.198 ± 0.417); SAM2-large: 0.065 (0.167 ± 0.257). These ΔE00 values lie well below the ≈1 just noticeable difference threshold on average, indicating close chromatic agreement between the predictions and annotations. Within a single dataset and training protocol, fine-tuned SAM2, especially its larger variants, provides robust spatial accuracy, boundary reliability, and color fidelity suitable for clinical shade-matching workflows, while UNet offers a competitive convolutional baseline. These results indicate technical feasibility rather than clinical validation; broader baselines and external, multi-center evaluations are needed to determine its suitability for routine shade-matching workflows. Full article

This study explored the Erasmus+ project ’H2OMap: Innovative Learning by Hydraulic Heritage Mapping’, integrating environmental awareness and cultural heritage into secondary education through interdisciplinary, ICT, and STEM-based approaches. Focused on water-related heritage in the Mediterranean, the study pursued three aims: integrate ICT-supported participatory mapping bridging history/geography subjects with digital innovation; identify learning benefits and implementation conditions; and generate transferable outputs and datasets for classroom reuse. Intellectual outputs include a methodological guide, an e-learning course, and an educational multiplatform comprising a mobile mapping app for in situ geocataloguing, an online database, and a geoportal with interactive StoryMaps. Evidence came from classroom testing across age groups, teacher feedback from the e-learning course, student mobilities in Spain, Italy, and Portugal, and platform usage records. More than 390 students and teachers participated, documenting over 100 hydraulic heritage elements. Additionally, dissemination through nine multiplier events and conferences reached over 550 external attendees. Findings show increased student engagement and ICT/GIS skills, clearer cross-curricular integration, and a replicable open workflow supported by structured coordination that strengthens school–university partnerships. Learner experience emphasised hands-on, place-based exploration and collaborative documentation of water heritage. Recommendations include using open geospatial standards, providing teacher training, and maintaining geoportals for classroom reuse. Full article

Unmanned aerial vehicles (UAVs) often face significant challenges in trajectory tracking within complex dynamic environments, where uncertainties, external disturbances, and nonlinear dynamics hinder accurate and stable control. To address this issue, a bio-inspired deep reinforcement learning (DRL) algorithm is proposed, integrating behavior cloning (BC) and long short-term memory (LSTM) networks. This method can achieve autonomous learning of high-precision control policy without establishing an accurate system dynamics model. Motivated by the memory and prediction functions of biological neural systems, an LSTM module is embedded into the policy network of the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. This structure captures temporal state patterns more effectively, enhancing adaptability to trajectory variations and resilience to delays or disturbances. Compared to memoryless networks, the LSTM-based design better replicates biological time-series processing, improving tracking stability and accuracy. In addition, behavior cloning is employed to pre-train the DRL policy using expert demonstrations, mimicking the way animals learn from observation. This biomimetic plausible initialization accelerates convergence by reducing inefficient early-stage exploration. By combining offline imitation with online learning, the TD3-LSTM-BC framework balances expert guidance and adaptive optimization, analogous to innate and experience-based learning in nature. Simulation experimental results confirm the superior robustness and tracking accuracy of the proposed method, demonstrating its potential as a control solution for autonomous UAVs. Full article

Artificial Intelligence and machine learning are increasingly used to interrogate complex biological data. This systematic review evaluates their application to multi-omics for the molecular characterization of hematological malignancies, an area with unmet clinical need. We searched PubMed, Embase, Institute of Electrical and Electronics Engineers Xplore, and Web of Science from January 2015 to December 2024. Two reviewers screened records, extracted data, and used a modified appraisal emphasizing explainability, performance, reproducibility, and ethics. From 2847 records, 89 studies met inclusion criteria. Studies focused on acute myeloid leukemia (34), acute lymphoblastic leukemia (23), and multiple myeloma (18). Other hematological diseases were less frequently studied. Methods included Support Vector Machines, Random Forests, and deep learning (28, 25, and 24 studies). Multi-omics integration was reported in 23 studies. External validation occurred in 31 studies, and explainability in 19. The median diagnostic area under the curve was 0.87 (interquartile range 0.81 to 0.94); deep learning reached 0.91 but offered the least explainability. Artificial Intelligence and machine learning show promise for molecular characterization, yet gaps in validation, interpretability, and standardization remain. Priorities include external validation, interpretable modeling, harmonized evaluation, and standardized reporting with shared benchmarks to enable safe, reproducible clinical translation. Full article

This paper introduces a novel Underclothing Body Shape Perception Network (UBSP-Net) for reconstructing parametric 3D human models from clothed full-body 3D scans, addressing the challenge of estimating body shape and pose beneath clothing. Our approach simultaneously predicts both the internal body point cloud and a reference point cloud for the SMPL model, with point-to-point correspondence, leveraging the external scan as an initial approximation to enhance the model’s stability and computational efficiency. By learning point offsets and incorporating body part label probabilities, the network achieves accurate internal body shape inference, enabling reliable Skinned Multi-Person Linear (SMPL) human body model registration. Furthermore, we optimize the SMPL+D human model parameters to reconstruct the clothed human model, accommodating common clothing types, such as T-shirts, shirts, and pants. Evaluated on the CAPE dataset, our method outperforms mainstream approaches, achieving significantly lower Chamfer distance errors and faster inference times. The proposed automated pipeline ensures accurate and efficient reconstruction, even with sparse or incomplete scans, and demonstrates robustness on real-world Thuman2.0 dataset scans. This work advances parametric human modeling by providing a scalable and privacy-preserving solution for applications to 3D shape analysis, virtual try-ons, and animation. Full article

Background/Objectives: This study presents Co.Ge. a Cognitive Generative digital platform for cognitive testing. We describe its architecture and report a pilot study. Methods: Co.Ge. is modular and web-based (Laravel-PHP, MySQL). It can be used to administer a variety of validated cognitive tests, facilitating administration and scoring while capturing Reaction Times (RTs), trial-level responses, audio, and other data. Co.Ge. includes a study-management dashboard, Application Programming Interfaces (APIs) for external integration, encryption, and customizable options. In this demonstrative pilot study, clinical and non-clinical participants completed an Auditory Verbal Learning Test (AVLT), which we analyzed using accuracy, number of recalled words, and reaction times as outcomes. We collected ratings of user experience with a standardized rating scale. Analyses included Frequentist and Bayesian Generalized Linear Mixed Models (GLMMs). Results: Mean ratings of user experience were all above 4/5, indicating high acceptability (n = 30). Pilot data from AVLT (n = 123, 60% clinical, 40% healthy) showed that Co.Ge. seamlessly provides standardized clinical ratings, accuracy, and RTs. Analyzing RTs with Bayesian GLMMs and Gamma distribution provided the best fit to data (Leave-One-Out Cross-Validation) and allowed to detect additional associations (e.g., education) otherwise unrecognized using simpler analyses. Conclusions: The prototype of Co.Ge. is technically robust and clinically precise, enabling the extraction of high-resolution behavioral data. Co.Ge. provides traditional clinical-oriented cognitive outcomes but also promotes complex generative models to explore individualized mechanisms of cognition. Thus, it will promote personalized profiling and digital phenotyping for precision psychiatry and rehabilitation. Full article

Berries are popular by consumers for improving vision, lowering blood sugar, improving circulation, and cardiovascular protection. They are usually small, thin-skinned, and fragile, with inconsistent ripening times. Harvesting robots are able to accurately determine the ripeness of fruits, avoiding pulp breakage and nutrient loss caused by manual squeezing. This work reviews the development and application of berry harvesting robots with market prospects in recent years. Next, this paper discusses the key technologies of berry picking robots, including fruit detection and localization technology, motion planning technology, and end-effector and harvesting mechanism. It also discusses the challenges currently faced in the development of berry harvesting robots, including external factors such as unstructured working environments and internal technical difficulties such as robot design and control. To address these challenges, future berry picking robots should focus on developing weak supervision recognition models based on deep learning, high-speed collision-free multi-arm collaborative harvesting technology, and high fault-tolerant harvesting technology to improve picking efficiency and quality, reduce fruit damage, and promote the automation and intelligence of the berry harvesting. Full article