A104: Explainable AI-Driven Multimodal Biomechanics Analysis of Cheerleading Rotational Skills

Expert consensus-derived evaluation criteria for orthodontic treatment outcomes using a novel ranking method: A retrospective dental cast analysis study.

In China, most quality assessments (QAs) of the high-contrast resolution of medical computed tomography (CT) equipment is performed by subjective evaluation, which is simple but dependent on observer expertise. The purpose of this study was to compare the accuracy and degree of dispersion of evaluation results from observers with different expertise and to develop an objective, easily understandable new approach to minimize the error caused by the subjective nature of observer evaluation. The accuracy and reproducibility of this new approach was validated and compared using both subjective and objective evaluation methods. This new objective evaluation method of high-contrast resolution was based on measurements of CT phantom images. The intensity profile of the aluminum line pairs' (LPs') images was extracted and quantified following the Rayleigh criterion. The interpolation method was employed to improve the consistency of the middle profile lines of each LP group for application of the Rayleigh criterion. With the introduction of a normalized margin, the accuracy of the results increased to the decimal level. Five junior observers and three QA professionals evaluated the high-contrast resolution of three different CT scanners using the subjective method, modulation transfer function (MTF) method, and the new objective method, and the results were compared statistically. The experimental standard deviations ranged from 0.4LP/cm to 0.55LP/cm for subjective visual inspection. The uncertainty of the expert group's results was much lower than that of the volunteer group. The experimental standard deviations of the MTF evaluation, according to both the standard deviation (SD) and point spread function (PSF) methods, were within the intervals of (0.04-0.06)LP/cm and (0.08-0.16)LP/cm, whereas that of the proposed objective evaluation method was approximately 0.02LP/cm. The experimental data indicated that the results of the new objective evaluation were in accordance with those of the expert group's subjective evaluation and the MTF evaluation. In this study, the subjective evaluation results from experienced observers showed relatively less uncertainty and more accuracy than those from inexperienced observers. Furthermore, a new objective evaluation approach that is easily understood, accurate, and reproducible was developed in this study. Results acquired using this new method were comparable with those of expert groups' subjective evaluation and MTF evaluation. Specialized QA software that integrates the steps implemented in the new method is being developed to enhance the convenience of the approach and assist observers in obtaining accurate and replicable assessment results.

<abstract><p>Muscle coordination and motor function of stroke patients are weakened by stroke-related motor impairments. Our earlier studies have determined alterations in inter-muscular coordination patterns (muscle synergies). However, the functional connectivity of these synergistically paired or unpaired muscles is still unclear in stroke patients. The goal of this study is to quantify the alterations of inter-muscular coherence (IMC) among upper extremity muscles that have been shown to be synergistically or non-synergistically activated in stroke survivors. In a three-dimensional isometric force matching task, surface EMG signals are collected from 6 age-matched, neurologically intact healthy subjects and 10 stroke patients, while the target force space is divided into 8 subspaces. According to the results of muscle synergy identification with non-negative matrix factorization algorithm, muscle pairs are classified as synergistic and non-synergistic. In both control and stroke groups, IMC is then calculated for all available muscle pairs. The results show that synergistic muscle pairs have higher coherence in both groups. Furthermore, anterior and middle deltoids, identified as synergistic muscles in both groups, exhibited significantly weaker IMC at alpha band in stroke patients. The anterior and posterior deltoids, identified as synergistic muscles only in stroke patients, revealed significantly higher IMC in stroke group at low gamma band. On the contrary, anterior deltoid and pectoralis major, identified as synergistic muscles in control group only, revealed significantly higher IMC in control group in alpha band. The results of muscle synergy and IMC analyses provide congruent and complementary information for investigating the mechanism that underlies post-stroke motor recovery.</p></abstract>

Subjective evaluations of athletic performance are an important part of decision making across sporting organisations. Based on their expertise and intuition, coaches select their starting line-ups, scouts recommend or discourage teams from signing new potential players, and academy directors decide which players should move up or out of a team’s academy system. While this intuitive evaluation of performance occurs constantly throughout sport, little attention has been given to how this process can be more formally designed, implemented, and assessed to capture these expert evaluations of performance more effectively and thereby better inform decision making within sports organisations. In this opinion paper, the discrepancy between how often subjective evaluations are used in practice and how little research exists regarding their development, implementation, and effectiveness is examined. To assist in moving the sport science field forward, this paper will attempt to define subjective performance evaluations, describe the potential benefits that sporting organisations may get from using them, detail existing and potential uses for these measures, provide a framework for understanding how subjective data collection, and call upon the academic and applied sport community to drive the subjective evaluation field further.

What's in the box?! Towards explainable machine learning applied to non-residential building smart meter classification

We employ multiple predictive algorithms combined with explainable machine learning techniques to forecast and interpret Airbnb rental prices in Sydney, Australia. The best-performing model is selected using multiple metrics and model confidence sets from a variety of methods ranging from simple linear regression to more complex forecast combinations. In addition, we evaluate the importance of feature engineering by training the models on datasets constructed with and without feature engineering and assessing their respective accuracies. Ensemble methods, particularly stacking regressions, outperform other algorithms on both the training and test datasets, while linear models perform the worst. Factors such as property capacity, proximity to popular areas and luxury amenities increase price predictions according to Shapley values, whereas being near major highway entrances is linked to lower prices, likely due to noise and air pollution.

BackgroundArthritis is a prevalent chronic disease substantially impacting patients’ quality of life (QoL). While identifying key determinants associated with arthritis is critical for targeted interventions, traditional statistical methods often struggle with complex interactions, and existing machine learning (ML) approaches frequently lack the interpretability needed to guide clinical decisions. This study integrates a comprehensive, explainable machine learning (XAI) workflow to identify and interpret key QoL-related predictors of arthritis status in a large national cohort.MethodsData were obtained from 15,011 participants aged > 45 years in the 2020 China Health and Retirement Longitudinal Study (CHARLS). We initially selected 55 potential QoL-related predictors spanning demographic, functional, pain, psychosocial, and lifestyle domains. Feature engineering was performed to create aggregate scores, indicators, and binned variables. Missing data were handled using imputation combined with missing indicator variables. A LightGBM-based feature selection process identified 68 key predictors. Nine ML models (including Logistic Regression, RandomForest, GradientBoosting, LightGBM, CatBoost, XGBoost, DecisionTree, NaiveBayes, KNN) were developed using SMOTE-resampled training data, with hyperparameters optimized via Optuna targeting recall. Performance was evaluated on a held-out test set using Area Under the ROC Curve (AUC), Average Precision (AP), Recall, Specificity, Precison, and F1-score. SHapley Additive exPlanations (SHAP) analysis was applied to the best-performing model (GradientBoosting) for interpretation.ResultsSeveral models achieved strong predictive performance, with GradientBoosting yielding the highest AUC (0.767, 95% CI: 0.752–0.782) and high AP (0.678, 95% CI: 0.655–0.702). SHAP analysis identified multi-site pain burden (particularly knee/leg pain and pain location count), age, self-rated health, sleep quality, functional limitations (ADL counts/scores), and negative affect as the most influential predictors driving arthritis status prediction.ConclusionsThis study successfully applied an XAI pipeline to identify and rank key QoL-related factors predictive of arthritis status in a large Chinese cohort, achieving robust model performance. Pain burden, age, subjective health, sleep, functional status, and psychological well-being are critical determinants. These interpretable findings can inform risk stratification and guide targeted interventions focusing on these key areas to potentially improve arthritis management.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12955-025-02412-9.

The COVID-19 pandemic poses new challenges on pharmaceutical supply chain including the delays and shortages of resources which lead to product backorders. Backorder is a common supply chain problem for pharmaceutical companies which affects inventory management and customer demand. A product is on backorder when the received quantity from the suppliers is less than the quantity ordered. Knowing when a product will be on backorder can help companies effectively plan their inventories, propose alternative solutions, schedule deliveries, and build trust with their customers. In this paper, we investigate two problems. One is how to use machine learning classifiers to predict product backorders for a pharmaceutical distributor. The second problem we focused on is what are the particular challenges and solutions for such task under a pandemic setting. This backorder dataset is different from existing benchmark backorder datasets with very limited features. We discuss our challenges for this task including empirical data pre-processing, feature engineering and understanding the special definitions of backorder under the pandemic situation. We discuss experimental design for predicting algorithm and comparison metrics, and demonstrate through experiments that decision tree algorithm outperforms other algorithms for our particular task. We show through explainable machine learning approaches how to interpret the prediction results.

Background Cystic Echinococcosis (CE) is a zoonotic parasitic disease caused by the larvae of Echinococcus granulosus. Despite global efforts in prevention and control, systematic studies on women of childbearing age(WCBA) remain relatively scarce. Methods The data for this study were obtained from the Global Burden of Disease (GBD) 2021 database, covering the epidemiological burden of CE in women of childbearing age(WCBA) globally from 1990 to 2021.The study used Estimated Annual Percentage Change (EAPC) to perform a descriptive analysis of the trends in the disease burden of CE in WCBA from 1990 to 2021.The framework was constructed using the Extreme Gradient Boosting(XGBoost), with feature engineering and gradient boosting iteration to analyze the trends in incidence, prevalence, and Disability-Adjusted Life Years (DALYs).The trends in disease burden from 2022 to 2050 were predicted, and the Shapley Additive Explanations (SHAP) values were used to quantify the contribution of features, identifying high-risk populations and regions. Results The study found that between 1990 and 2021, the global prevalence of CE increased by about 0.84 times, the incidence rose by about 0.88 times, while DALYs decreased by about 0.22 times.In terms of age distribution, the highest increase in CE prevalence among WCBA was seen in the 45–49 age group, with an increase of nearly 1.44 times; while the highest increase in incidence was observed in the 30–34 age group, with an increase of about 1.04 times.Regionally, the CE disease burden is heaviest in Low SDI regions, while the number of cases in Middle SDI regions increased by 1.36 times.According to the model predictions, the prevalence and incidence of CE will show some fluctuation trends in the future. It is expected that by 2050, the global prevalence of CE in WCBA will reach 10.13 per 100,000, and the incidence will reach 2.34 per 100,000. Conclusions The global burden of CE in WCBA remains high, showing regional and age-related heterogeneity. There is a need to strengthen prevention and control measures in resource-poor areas, optimize dog management and hygiene intervention strategies, to reduce maternal and child health risks.

Effects of peripheral fatigue of ankle evertors on force sense and electromyographic activity in healthy subjects.

Physique Related Perceptions and Biological Correlates of Eating Disorder Risk among Female Collegiate Equestrians Sport-specific features of certain sports are implicated as perpetuating eating disorder risk and this is especially prominent among female athletes competing in aesthetic sports. Form-fitting uniforms and where there is a subjective evaluation component to determining performance success characterize aesthetic sports.Contextual mechanisms common within this sport classification involve public weigh-ins, comments about physique, costume features and subjective evaluation by judges magnify the typical pressures associated with pursuing success in sport especially if physical characteristics do not meet sport demands.

To compare the assessment of serial visual fields (VFs) based on subjective expert evaluation with the fast and slow VF component rates determined with pointwise exponential regression (PER) and pointwise linear regression (PLR). A total of 5272 VF examinations from 376 eyes diagnosed with open-angle glaucoma were included. Three glaucoma specialists assessed each VF qualitatively to evaluate progression status and the qualitative rate of progression. The rates of VF decay were determined with PER and PLR at each VF location, which were ranked according to the regression coefficient and partitioned into 2 groups (fast and slow). A mean rate for the fast and slow partitions was obtained based on the average of the regression coefficients in each partition. κ-values were used to measure the agreement among the experts and the PER and PLR algorithms. The average baseline VF mean deviation for the study sample was -6.6 (±5.9) dB. The agreement of the likelihood of progression among the dichotomized experts' score and PER was moderate (κ=0.41, P<0.01) and fair (κ=0.39, P<0.01) for PLR. The agreement of the likelihood of progression among the 3 dichotomized experts' scores was fair (κ=0.22, P<0.01). The agreement of the area of worsening among the dichotomized experts' score and PER and PLR were both moderate (κ=0.48, P<0.01; κ=0.46, P<0.01). The eyes flagged by experts as having "fast" progression rates had a higher average rates of decay for PER and PLR at -2.7 (±4.1) %/year and -0.8 (±1.2) dB/year; eyes flagged as "slow" had lower rates of decay at -0.3 (±1.5) %/year and -0.1 (±0.5) dB/year. Expert qualitative evaluation of field series for change and rate of change correlate more closely with the fast component than with the slow component of VF decay.

Enhancing FRP-concrete interface bearing capacity prediction with explainable machine learning: A feature engineering approach and SHAP analysis

Wine quality assessment in the modern viticulture industry faces significant challenges due to reliance on subjective expert evaluation, leading to inconsistencies, scalability limitations, and high costs in commercial applications. Traditional approaches lack standardization and struggle with large-scale quality control processes. This comprehensive study evaluates the effectiveness of advanced machine learning techniques for wine quality prediction using a dataset of 6,497 samples (1,599 red wine and 4,898 white wine)0.14 original physicochemical features that were expanded to 34 through feature engineering, plus the quality target variable and addressed class imbalance using SMOTE. Our methodology integrates traditional algorithms (Logistic Regression, SVM, KNN), ensemble methods (Random Forest, XGBoost, LightGBM, Voting and Stacking ensembles), deep neural networks, and a novel application of transfer learning from white wine quality models to enhance red wine quality prediction. Results demonstrate superior performance of ensemble methods across evaluation metrics, with Random Forest achieving up to 95% accuracy and 0.994 AUC score in optimized configurations, while Voting and Stacking ensembles consistently delivered robust performance (81.5% accuracy, 85.3% F1 score) across varied testing conditions. Feature importance analysis using SHAP revealed that alcohol content, sulfur dioxide levels, and volatile acidity are the most influential predictors, with complex interaction patterns between chemical properties. Transfer learning showed promising results with faster convergence but slightly lower accuracy than models trained from scratch. This research advances both the methodological framework for wine quality prediction and provides actionable insights for the wine industry by identifying critical chemical determinants of wine quality.

Measuring the efficiency of athletes during competition has been a subject of interest both for experts and scientists in sports for more than a hundred years. Basketball has recognized in the 1940s how important it is to analyze efficiency indicators because these procedures allow coaches to increase their knowledge. There are two basic methods – objective and subjective – for evaluating the efficiency, or real quality of basketball players. The aim of this research is to establish the level of correlation between these two methods and to identify clusters, i.e. player hierarchy based on the results of both methods of efficiency evaluation. The sample of variables consisted of 12 basketball players who participated in the 2010 FIBA World Championships in Turkey. The subjective evaluation, also called expert evaluation, was performed by coaches of seven national teams that participated in the Championship. The objective evaluation was performed using the EEF efficiency index. The data was processed using z-scoring, the Pearson coefficient, and hierarchical cluster analysis. The Pearson coefficients of linear correlation between the efficiency index and the expert evaluation is r = 0.859 with a statistical significance of p ≤ 0.01. The cluster analysis distinguished two groups of players, which were named quality and super quality. The variance analysis showed that the probability of the clusters being equal is less than p ≤ 0.00. The research has shown that the evaluation by coaches is relevant and is fully consistent with the efficiency index formula. Also, the distinction of two groups of players by clustering is not uncommon in the basketball practice and is linked with efficiency at the given time.