Machine Learning Models for Identifying Dental Pain in Adolescents.

Recent advances in artificial intelligence (AI) are revolutionizing materials science by unlocking unprecedented capabilities in designing novel compounds and accurately predicting their properties. Among these, graph-based machine learning (ML) algorithms have garnered significant attention for their ability to capture complex atomic interactions and use them as effective descriptors. In this study, we integrated state-of-the-art generative AI (Gen AI) and ML techniques with quantum mechanical calculations to discover novel next-generation electrolytes for alkali metal batteries. We developed a Generative Adversarial Network (GAN) framework incorporating a graph-based generator and discriminator models to generate novel electrolyte candidates. The GAN model was trained on a subset of approximately 1 million molecules from the GDB-11 database, which enabled the generation of 30,000 unique and chemically valid molecules. Concurrently, a Message Passing Neural Network (MPNN) model was trained for property prediction by utilizing the QM9 dataset. Using the trained MPNN model, we predicted the properties of the newly generated molecules and screened the candidates based on the criteria of negative standard enthalpy of formation and a wide HOMO-LUMO gap. First-principles density functional theory (DFT) calculations were conducted for additional screening and to evaluate key thermodynamic and electrochemical properties, including standard enthalpy of formation, oxidation potential, and reduction potentials. Finally, a set of 26 promising candidates was acquired with outstanding electrochemical characteristics. Our findings demonstrate the potential of AI-driven approaches to discover high-performance, stable, and efficient electrolytes as promising alternatives to conventional organic electrolytes for next-generation energy storage systems.

Synthesizing T2-weighted MRI from CT scans presents a challenging ill-posed problem that remains underexplored in abdominopelvic imaging. We aim to develop and compare deep learning algorithms for generating synthetic T2-weighted MRI from pelvic CT, systematically evaluating architecture and training strategies for feasibility and performance.The framework adopts a conditional Generative Adversarial Network (GAN) approach. Three state-of-the-art models [efficient Self-Attention UNet (ESAUNet), Residual Vision Transformer (ResViT), and Cascaded Gaze] were utilized as generators. A combined loss function (L1, VGG19 perceptual, adversarial) was employed to optimize fidelity. A multicenter cohort (n = 90) including the SynthRad2023 (n = 60) and in-house (n = 30) datasets was used for model training. Five-fold cross-validation was used, and independent testing was performed on the Gold Atlas male cohort (n = 19). Blinded qualitative assessment by two radiologists was conducted. Rigorous qualitative and quantitative assessments were performed.On the held-out cohort, ESAUNet achieved the highest average performance across all metrics: PSNR 22.21dB, SSIM 0.748, and MAE 0.044. This surpassed the performance of the ResViT model (PSNR 21.18dB, SSIM 0.727) and CascadedGazeNet (PSNR 21.41dB, SSIM 0.703). Radiologists found no significant difference in overall pelvic image quality (p = 0.294) or rectal wall delineation (p = 0.358) between synthetic T2-weighted MRI and true MRI, despite lower ratings for fine details. Inter- and intra-observer agreement was robust.Conditional GAN architectures demonstrate technical feasibility for CT-to-T2-weighted pelvic MRI translation. ESAUNet's efficiency and robust performance highlight its potential as a key enabler for MRI-equivalent imaging in resource-limited settings.

Mapping reaction pathways on complex potential energy surfaces (PESs) and locating transition states (TSs) is often used for understanding chemical reaction mechanism(s). The nudged elastic band (NEB) method is widely used for this purpose, but it becomes computationally expensive for large systems due to the repeated evaluation of energies and forces. We present an active learning algorithm coupled with the nudged elastic band, AL-NEB, for efficient convergence to the TS. AL-NEB constructs a surrogate PES and actively selects training points in two phases: (a) Exploration-Exploitation and (b) Renunciation. Strategies have been introduced for making the algorithm efficient and stable. We show the efficacy of the algorithm on several 2D analytical potentials, HCN isomerization, keto-enol tautomerization, and high-dimensional heptamer island diffusion (up to 525 degrees of freedom). In all cases, AL-NEB locates the "exact" TS on the chosen model chemistry with an order-of-magnitude fewer force evaluations than the standard NEB, demonstrating its scalability and efficiency.

Development and Validation of a Deep Learning System for Echocardiographic Assessment of 16-Segment LV Wall Thickness.

The increasing demand for personalized healthcare solutions highlights the limitations of one-size-fits-all treatment strategies. Digital twin (DT) technology, which enables real-time virtual replicas of physical systems, offers a promising approach to advance personalized medicine through continuous monitoring, simulation, and prediction. This study presents the foundational phase of a machine-learning-driven biosensor DT designed to support personalized infertility treatment through integration with a Smart Health Monitoring System. The DT replicates the behavior of a field-effect transistor (FET)-based biosensor functionalized with 17$\beta$-estradiol aptamers and trained on experimental data obtained from silicon nanonet BioFET prototypes. Seven supervised machine learning algorithms were evaluated to predict hormone concentration from electrical parameters ($V_{g}$, $I_{sd}$). The K-Nearest Neighbors (KNN) model achieved the highest predictive accuracy ($R^{2} = 0.99$, $\text{CV}\text{-}R^{2} = 0.98$, RMSE = 11.87 pg/mL) and demonstrated robust cross-device generalization under Leave-One-Biosensor-Out validation ($R^{2} = 0.59$). These results confirm the model's capability to capture nonlinear relationships and generalize across independently fabricated sensors. The developed model constitutes a validated predictive core of a biosensor digital twin. At the current stage, the DT is limited to predictive modeling and does not yet implement real-time synchronization or closed-loop feedback, which are planned in future work. This study establishes a practical framework for data-driven digital twins of biosensors and demonstrates their potential for integration into smart health monitoring systems supporting personalized infertility care. The proposed approach provides a foundation for real-time, adaptive, and clinically relevant biosensor twins in precision medicine.

Drug-resistant epilepsy (DRE) affects approximately one-third of patients with epilepsy. The molecular heterogeneity underlying DRE remains poorly defined, largely due to limited access to resected brain tissue and substantial genetic diversity. Current classifications rely primarily on clinical symptoms and histopathological features rather than molecular mechanisms, constraining mechanistic insight and the development of targeted therapies. This study aimed to develop a transcriptome-based, machine learning-guided framework for molecular classification of DRE. We performed comprehensive RNA sequencing on 153 surgically resected samples from 95 patients with DRE. Two transcriptomic subtypes were identified through unsupervised clustering. We also leveraged a weighted correlation network-based framework and systematic transcriptional signature comparison and developed a classification model using machine learning algorithms. Unsupervised clustering revealed two molecular subtypes that diverged from traditional pathological classifications, indicating an alternative transcriptomic basis for epilepsy pathogenesis. A classification model was constructed based on four key differentially regulated pathways: (1) neuroactive ligand-receptor interaction, (2) cAMP signaling, (3) γ-aminobutyric acid (GABA)ergic synapse, and (4) calcium signaling. Among the tested algorithms, the random forest model demonstrated superior performance, achieving 96% classification accuracy with an area under the curve (AUC) of .95. These molecular subtypes and their pathways could serve as key molecular hallmarks of epilepsy, offering valuable insights for developing targeted therapies. Moreover, our findings introduce a novel framework for classifying epilepsy based on its molecular nature, potentially connecting the clinical symptoms with the underlying causes more effectively.

The global education system comprises many technical and non-technical institutions. The selection of an institute plays a very important role in shaping the career of a student. With such a massive number of choices out there, the decision of which institution to go to will be a huge challenge for parents as well as students. Unexpected events such as the Covid-19 pandemic even disrupted global higher education highlighting infrastructural gaps and pedagogical limitations in knowledge delivery through sudden transitions to remote learning. Institutions were financially unstable with reductions in enrolment especially of international students. Increased operations cost for digital infrastructure and health protocols also took a toll on the academia, and it became important to predict the position of the institutions effectively. Our research proposes a fuzzy time series based ensemble model, ensemble based time series association (EBTsA) for dynamically predicting institutional rankings in the highly uncertain academic environment. The model integrates a fuzzy time series and ensemble machine learning algorithm for institutional rank prediction and capturing inherent variations induced by ranking uncertainties. It uses the method of fuzzification to adaptively consider the importance of rankings in a changing way over time, both before and after the pre- and post-COVID changes. This vital rank gap in earlier studies has personified rankings as static or uniform. Various algorithms such as FTS, FCA, IFS, IFS_New and the proposed algorithm (EBTsA) are compared based on their performance in the dynamic ranking prediction. The EBTsA model quantifies ranking uncertainties and forecasts institutional ranks with a mean absolute percentage error (MAPE) of 7.12, a mean absolute scaled error (MASE) of 0.32, and a directional accuracy (DA) of 82.2, outperforming conventional deterministic models. The predictive performance of the model ensures highly accurate and reliable dynamic rank forecasts, enabling stakeholders to make informed decisions about educational institutions. Our study may contribute to two sustainable development goals (SDGs)of the United Nations Organisation (UNO), such as (SDG 4), which provides "quality education and its connection to inequality", and (SDG 10) for "reduced inequalities and its connection to education".

Miscarriage occurs in approximately 15% of all pregnancies, and recent studies have suggested a potential role of the microbiome. A nested case-control study from the Swedish Maternal Microbiome cohort was conducted, including 34 participants who sent at least one vaginal or fecal microbiome sample and questionnaire data before miscarrying (n = 34), and matched controls (n = 105 for regression models, n = 27 for machine learning models). Non-vaccine type HPV (aOR 3.95, 95%CI 1.04-15.06) and vaginal microbiome with community state type (CST) II (aOR 6.52, 95%CI 1.58-26.98) or CST-IVB (aOR 4.18, 95%CI 1.08-16.18) in early pregnancy were associated with an increased risk of miscarriage. Furthermore, we explored six machine learning algorithms using 70% of the cohort for training and 30% for testing, for the prediction of miscarriage using vaginal (AUROC 85%), fecal (AUROC 81%) and questionnaire (AUROC 82%) data separately and combined (AUROC 82%). Our results highlight the urgency of HPV screening and vaccine development for women's reproductive health. Despite limitations, including a small number of miscarriage cases, our results indicate the potential for both vaginal and fecal microbiomes in the prediction of miscarriage.

Epidemiological cohorts often collect self-reported oral health (SROH) questionnaires but lack clinical periodontal measurements. We developed a selective, explainable machine learning (ML) pipeline that can assign labels for severe periodontitis (SP) or no periodontitis (NP). Three datasets (n = 498) with SROH questionnaires, demographics, and Community Periodontal Index of Treatment Needs (CPITN) scores were used to derive NP, moderate periodontitis (MP), and SP categories. MP cases were excluded from model development. After cleaning and feature engineering, non-similar label duplicates were removed. A CatBoost model (Separator-A) was trained with tenfold cross-validation; NP/SP predictions were retained when probability ≥ 0.85. From these outputs and domain rules, a rule-consistent subset was created to train a second model (Separator-Z). Performance was evaluated on internal test and hold-out inference sets. Next, the pipeline was applied to MP cases. The pipeline achieved complete separation across all evaluation sets within the retained high-confidence subset, which represented 4.31% of eligible NP/SP cases, while no MP cases were misclassified as NP or SP. Thus, a two-stage, explainable ML pipeline can selectively identify SP and NP from SROH questionnaire data, supporting case-control selection in cohorts without clinical periodontal examinations, though validation is warranted to confirm generalizability.

To develop and internally test a multiparametric radiomics combined model for differentiating benign and malignant orbital tumors. This retrospective study analyzed 147 patients from two centers (December 2014 to March 2024) with pathologically confirmed orbital tumors and preoperative contrast-enhanced magnetic resonance imaging(MRI). After image preprocessing, 3668 radiomics features were extracted from T2-weighted imaging (T2WI) and contrast-enhanced T1-weighted imaging (CE T1WI) sequences. Feature reduction and selection were performed using the t-test/U-test, Pearson correlation coefficient, minimum redundancy maximum relevance (mRMR), and least absolute shrinkage and selection operator (LASSO) regression. Three machine learning algorithms, logistic regression (LR), naive Bayes classifier (NaiveBayes), and Multilayer perceptron (MLP) were used to construct radiomics models. A combined radiomics model (CRM), defined as an MLP-based model incorporating selected features from both T2WI and CE T1WI sequences, was subsequently built and integrated with clinical factors to create a radiomics nomogram. Model performance was evaluated using the area under the curve (AUC), accuracy, sensitivity, and specificity. Decision curve analysis (DCA) assessed clinical utility, and SHapley Additive exPlanations (SHAP) provided model interpretability. Six key radiomics features were selected to establish the CRM. The MLP-based model achieved the highest AUC among the individual machine learning models in both training and test cohorts. The CRM demonstrated superior performance compared to models based solely on T2WI or CE T1WI, with AUCs of 0.877 (training cohort) and 0.860 (test cohort). The final nomogram, integrating the CRM and clinical factors, showed favorable discriminatory performance, achieving AUCs of 0.890 and 0.846 in the training and test cohorts, respectively. SHAP analysis identified 'squareroot_firstorder_Skewness_CE T1WI' and 'wavelet_LLH_glcm_Correlation_CE T1WI' as important predictors for malignant orbital tumors. This study presents an effective and explainable multiparametric MRI radiomics model that accurately differentiates benign from malignant orbital tumors. The developed nomogram demonstrates promising performance within the internal validation framework and may provide supportive information for clinical decision-making pending further external validation.

Gastric cancer (GC) risk is shaped by environmental exposures such as benzo[a]pyrene (BaP). Here, we systematically identified BaP-toxicological targets and dissected their contribution to GC development. BaP-related targets were independently predicted with stringent filters from ChEMBL, Similarity Ensemble Approach (SEA) and PharmMapper databases, while GC-related targets were mined from the Comparative Toxicogenomics Database (CTD), GeneCards and OMIM databases. Overlapping targets were subjected to protein-protein interaction (PPI) network construction, functional enrichment analysis and molecular docking. We then integrated multi-omics data using ten clustering algorithms to identify the consensus GC subtypes, which were subsequently employed 101 machine learning combinations to develop a consensus benzo[a]pyrene-related signature (CBRS) for GC patients. As a result, we identified seven hub toxicological targets: ALB, HSP90AA1, ESR1, INS, TP53, TNF, and EGFR, underscoring their potential central roles in BaP-driven GC pathogenesis. These targets are enriched in the MAPK, Lipid and atherosclerosis, and PI3K-Akt signaling pathway. The BaP-toxicological classifiers and the CBRS prognostic model could provide useful support for risk stratification and inform personalized therapeutic strategies for GC patients. Molecular docking results suggest that BaP exhibits relatively strong binding affinity with these key toxicological targets, potentially implicating their involvement in BaP-induced gastric cancer toxicity. Therefore, this study integrates multi-dimensional omics data with advanced machine learning algorithms to establish a comprehensive analytical framework for the toxicological effects of between BaP and GC, which transcends the limitations of traditional analyses and offers unprecedented insights and evidence chains for elucidating the pathogenesis of GC.

Primary open-angle glaucoma (POAG) is the leading cause of irreversible blindness. Regrettably, the roles of ferroptosis-related (FR) genes in POAG remain elusive. Five GEO data sets and a series of experimentations in vitro were used for bioinformatic exploration and biological validation. Using multiple machine learning algorithms, four critical FR genes in POAG progression were screened. The clinical value and biological function of NOX1 were comprehensively analyzed using bioinformatic approaches. A POAG in vitro model was constructed using H2O2 treatment. NOX1 effects on the viability of retinal ganglion cells (RGCs) and ferroptosis process were determined through CCK8, EdU, ROS detection, and transmission electron microscopy. Its upstream transcriptional mechanisms were determined through dual luciferase assays, and chromatin immunoprecipitation (ChIP). NOX1 was identified as the critical FR gene in POAG progression and served as an effective diagnostic biomarker. High-NOX1 expression was tightly associated with increased infiltration levels of multiple subtypes of T cells, such as T cells CD8 and T cells CD4. However, the enrichments of eight metabolic gene sets did not differ between the POAG samples with high- and low-NOX1 expression groups. Silencing NOX1 maintained RGC survival and inhibited the ferroptosis process. Mechanistically, STAT3 upregulated NOX1 by binding its promoter region that was located at the 429th to 419th bases upstream of the NOX1 transcriptional start site. NOX1 overexpression reversed the inhibitory effects of silencing STAT3 on RGC survival and the ferroptosis process. NOX1 was a good biomarker for characterizing POAG and promoted POAG progression through STAT3-mediated transcriptionally activation.

Predictors of self-reported sexually transmitted infections (STIs) among men in 54 low and middle income countries (LMICs): a comparison of deep learning and classical machine learning algorithms.

Cancer remains a major global health challenge due to its genetic variability and intricate molecular mechanisms, which complicate the development of effective therapies. This review elucidates the integration of AI-driven methodologies in nanoparticle (NP) design, optimizing drug delivery systems (DDSs) for targeted cancer therapy. AI's predictive analytics facilitate the rational design of nanocarriers, enhancing drug bioavailability, optimizing pharmacokinetics, and improving tumor penetration. The incorporation of machine learning (ML) models accelerates NP fabrication, enabling real-time simulation of tumor dynamics and drug release kinetics. Furthermore, AI-powered platforms, such as EVOnano, simulate in silico tumor microenvironments to refine nanocarrier functionalities. This synergy fosters the development of next-generation smart therapeutics, wherein adaptive nanomedicines exhibit enhanced tumor specificity while mitigating systemic toxicity. Challenges, such as nanoparticle scalability, AI interpretability, and biological heterogeneity, persist, necessitating interdisciplinary advances. Nevertheless, AI-assisted nanomedicine signifies a paradigm shift towards highly efficacious, patient-tailored cancer interventions, revolutionizing treatment landscapes and propelling oncology into a new frontier of data-driven, precision-based therapeutics.

Individuals with cardiovascular-kidney-metabolic (CKM) syndrome exhibit a substantially elevated risk of all-cause and cardiovascular-specific mortality. Although estimated glucose disposal rate (eGDR) and a body shape index (ABSI) are commonly used indicators for assessing insulin resistance and atherosclerotic risk, respectively, evidence regarding their combined effect on all-cause and cardiovascular-specific mortality in patients with CKM syndrome remains insufficient. Investigating this combined impact may help improve risk stratification in this population. This study utilized data from the National Health and Nutrition Examination Survey (NHANES, 1999-2018), including 18,186 individuals with stage 0-4 CKM syndrome. Cox proportional hazards models, Kaplan-Meier curves and subgroup analyses were used to evaluate the associations between eGDR and ABSI and mortality risk. The integrated discrimination improvement (IDI) and net reclassification index (NRI) were used to assess the incremental prognostic value of eGDR and ABSI. Finally, six machine learning algorithms were applied to develop predictive models. During the follow-up period, a total of 2536 all-cause mortality and 790 cardiovascular-specific mortality were documented. After multivariable adjustment, both low eGDR and high ABSI independently predicted mortality risk. Combined analysis revealed that individuals with both Low-eGDR and High-ABSI had the highest mortality risk: all-cause mortality hazard ratio (HR) = 2.79 (95% CI 2.30-3.38) and cardiovascular-specific mortality HR = 4.53 (95% CI 2.96-6.92). However, the interaction effect was not statistically significant. Among the six machine learning algorithms, XGBoost demonstrated the best performance, with areas under the curve (AUC) of 0.877 and 0.850 for predicting all-cause and cardiovascular-specific mortality, respectively. Both eGDR and ABSI are independent and combined predictors of mortality risk among individuals with CKM syndrome. Their combined use significantly improves risk stratification and machine learning models provide an effective tool for precise risk assessment in this population.

Glucose predictions aim to empower continuous glucose monitoring (CGM) users by enabling preventive actions to reduce adverse glycemic events. The Accu-Chek® SmartGuide Predict app offers several AI-enabled predictive features, driven by machine learning algorithms. These include notifications for a low glucose predict within 30 min (LGP) and for nighttime low glucose risk, as well as a 2-h continuous glucose forecast. This study aimed to quantify the potential glycemic benefits of using the Predict app's predictive features in an adult population with type 1 diabetes (T1D). A comparative in silico study was conducted using the clinically backed University of Virginia Replay digital twin simulator. A control arm, simulating standard hypoglycemia and hyperglycemia mitigation strategies in line with international guidelines, was compared against intervention arms that incorporated probabilistic user behavior models responding to the app's predictive features. The evaluation was performed on 204 digital twins, representing 29,929 days of data, generated from the REPLACE-BG clinical trial dataset. Results demonstrated that using the app's predictive features has the potential to improve glycemic control in adults with T1D. The simulated intervention led to an average 2.9 percentage point reduction in time below range (<70 mg/dL), and a clinically significant increase of more than 3.6 percentage points in time in range (70-180 mg/dL). Furthermore, the daily number of CGM hypoglycemia alarms (<70 mg/dL) was reduced by 67%. The findings also suggest that consuming 10 g of fast-acting carbohydrates in response to LGP notifications provides an optimal balance, effectively preventing hypoglycemia while limiting rebound hyperglycemia. This in silico evaluation provides strong evidence supporting the potential clinical utility of the Accu-Chek SmartGuide Predict app for improving glycemic management in adults with T1D.

Abstract - The identification, segmentation, and extraction of tumor-affected regions from Magnetic Resonance Imaging (MRI) scans are important tasks in medical diagnosis. These processes are usually performed by radiologists or medical experts and require significant time and experience. Image processing techniques help in visualizing the anatomical structures of human organs more effectively, but detecting abnormal structures in the brain using basic imaging methods is still difficult. In this study, a Fully Automatic Heterogeneous Segmentation using Support Vector Machine (FAHS-SVM) method is proposed for brain tumor segmentation using deep learning techniques. The approach introduces a fully automated algorithm that separates the cerebral venous system in MRI images by utilizing structural, morphological, and relaxometry information. The segmentation process ensures a high level of consistency between the brain anatomy and the surrounding tissues. Extreme Learning Machine (ELM), which contains one or more hidden layers, is applied as a learning algorithm and is commonly used for tasks such as regression and classification. In this work, a probabilistic neural network classifier is used to train and evaluate the detection accuracy of tumors in brain MRI images. Experimental results demonstrate that the proposed system can effectively distinguish between normal and abnormal brain tissues with an accuracy of approximately 98.51%, showing the effectiveness of the proposed method.

The food industry is witnessing the emergence of specialized protein-based functional ingredients for the use as gelling, thickening, and/or emulsifying agents in various food applications. Different sources of protein including species and cultivars, as well as variable processing conditions affect the protein's structural characteristics, which in turn govern their functional properties. The complex relationship between the structure and function of the protein can be modeled using machine learning (ML) algorithms. In this study, different ML algorithms were used to predict solubility, emulsifying activity index, emulsifying capacity, and gel strength of different plant proteins using structural predictors (surface hydrophobicity, zeta potential, undenatured protein content, water holding capacity, soluble protein polymer content, β-sheet content). Model performances were assessed by specific metrics ( , mean absolute error [ ], and root mean squared error [ ]) and non-violation of physical constraints. The solubility and emulsifying activity index were predicted using surface hydrophobicity, zeta potential, and undenatured protein content. Emulsifying capacity was predicted using surface hydrophobicity, solubility, undenatured protein content, while gel strength was predicted using solubility, undenatured protein content, water holding capacity, soluble protein polymer content, and β-sheet content. The based Support Vector Regression model accurately predicted solubility ( = 0.8906), emulsifying activity index ( = 0.7383), emulsifying capacity ( = 0.7978), and gel strength ( = 0.8822). Results highlighted the potential of ML algorithms for predicting of plant protein functionality using a few macromolecular structural characteristics. Such predictive models could serve as indispensable tools in the selection of protein ingredients for various food applications.

Automated Measurement of Horizontal Strabismus in Children's Primary Gaze Photographs Using Deep Learning and Computer Vision.