University of Sousse

• Identifying the type and freshness level of beef where beef is primarily sold in packaged form. • Impedance spectroscopy enables non-invasive meat assessment, existing methods often rely on raw data or complex feature sets, failing to explain the rationale behind their selection or justify their complexity. • A systematic, physics-based feature engineering methodology that can classify muscle type and freshness in bovine meat simultaneously using a single neural network. • Feature selection is primarily based on significant parameter variation measured over a 14-day period, enabling a classification accuracy of up to 91,5% using both neural networks and SVM models. • The evolution of these parameters aligns with biological fundamentals, confirming the relevance of the selected features and underlying physical principles In this paper, we propose a systematic, physics-based feature engineering methodology that can classify muscle type and freshness in bovine meat simultaneously using a single neural network. We identify a minimal, yet highly informative set of features extracted from bioimpedance spectra based on observed spectral behavior and underlying physical principles. Feature selection in this study is primarily based on significant parameter variation measured over a 14-day period, particularly between days 2 or 3 and day 14 for a binary classification of meat in the two classes safe-to-sell and not-safe-to-sell. This variation serves as a strong indicator of sensitivity and contributes to achieving a high classification rate. The key features selected include parameters from the modified Fricke model and characteristic points in the α and β frequency bands (Rex, Rin, C, Im1, Im2 and Im3). We validate the stability of these features and assess the model’s interpretability. Our analysis highlights the consistent relevance of these parameters, enabling a classification accuracy of up to 91,5% using both neural networks and SVM models. The evolution of these parameters aligns with biological fundamentals, confirming the relevance of the selected features.

Brain-computer interfaces (BCIs) were initially created to help individuals with disabilities control devices and communicate without muscle movement. Today, BCIs are used for prosthetic control, cognitive enhancement, and neurological rehabilitation. The BCI system depends on analyzing electroencephalogram (EEG) signals captured from the brain. Decoding these EEG signals is a complex process that combines multiple algorithms to extract meaningful information from these intricate and noisy signals. One of the most popular techniques is the Common Spatial Patterns (CSP), which helps preserve useful and sensitive information. This paper presents an optimized extension of the CSP model for extracting EEG data features in a multiclass setting using Riemannian geometry-based weighting. The use of weighting based on Riemannian geometry enhances the robustness of covariance matrix computation, thereby decreasing the influence of noise that can significantly distort the mean of covariance matrices in the traditional CSP method. The proposed approach is also extended by the integration of a multi-band filter bank, providing a more detailed examination of EEG signals. Three classifiers, Linear Discriminant Analysis (LDA), Random Forest Classifier (RFC), and Multi-Layer Perceptron (MLP), are employed to differentiate features across four motor imagery tasks. LDA achieves an accuracy of 80.40%, while MLP and RFC reach 80.02% and 80.90%, respectively. The results obtained using a majority vote combining the decisions of the three classifiers are 81.83% for accuracy and Recall, 82.74% for precision, and 81.87% for F1-score. The proposed architecture is evaluated using the BCI Competition IV set 2a dataset, proving its effectiveness in EEG signal classification for BCI applications.

Abstract We consider matrix orthogonal polynomials related to Bessel type matrices of weights that can be defined in terms of a given matrix Pearson equation. From a Riemann–Hilbert problem we derive first and second order differential relations for the matrix orthogonal polynomials and functions of second kind. It is shown that the corresponding matrix recurrence coefficients satisfy a non-Abelian extension of a family of discrete Painlevé d-PIV equations. We present some nontrivial examples of matrix orthogonal polynomials of Bessel type.

This study explores the application of predictive data science techniques to anticipate customer behavior in the mobile telecommunications sector, integrating insights from finance and marketing. Leveraging advanced predictive models, including a multitask learning approach, the research is supported by an interactive web interface featuring a home- page, a Power BI dashboard, and a prediction page. The aim is to transform raw historical customer data into action- able insights for marketing teams, enabling accurate forecasts of mobile internet package activation and customers’ potential future value. Findings indicate that customer satisfaction and perceived sustainable value significantly influence subscription and recharge decisions, thereby enhancing loyalty and revenue generation. By emphasizing the synergy be- tween business needs, technological tools, and methodological frameworks, this work offers an innovative combination of theoretical and empirical approaches to advance practices within the telecommunications industry. Future research directions include incorporating real-time data streams and developing automated marketing recommendations to further optimize strategic effectiveness.

Microplastics (MPs) and perfluorooctane sulfonic acid (PFOS) are emerging pollutants that can co-occur in agricultural soils, but their combined impacts on crops are not yet fully understood. We investigated early-stage responses of broad bean (Vicia faba) grown for 28 days in greenhouse soil exposed to environmental MPs (50 mg kg⁻¹), PFOS (10 or 100 µg kg⁻¹), and their combinations. Targeted high-resolution mass spectrometry analyses quantified PFOS in plant tissues and glutathione redox status (GSH/GSSG), whereas untargeted metabolomics investigated global biochemical changes. Individually, MPs and PFOS reduced shoot length (29-33 % vs control). Conversely, biomass increased under co-exposure, indicating a distinct physiological adjustment. Noteworthy, PFOS accumulated dose-dependently and was enhanced by MPs (up to 6.92 ng g⁻¹ FW in shoots and 263.36 ng g⁻¹ FW in roots), evidencing MP-mediated modulation of PFOS bioavailability. Redox balance shifted toward oxidation: GSH/GSSG declined across treatments, especially in roots at high PFOS (-60 %). Metabolomics, followed by multivariate statistical analysis, successfully separated the control from the treated plants, with AMOPLS apportioning variance primarily to PFOS and the PFOS×MPs interaction. Discriminant metabolites in shoots highlighted a dose-related accumulation of isoprenoid and fatty-acid biosynthesis, stronger under MPs+high PFOS, while roots exhibited suppressed isoprenoids and enhanced phenylpropanoids at high co-exposure. Our results highlighted an MPs-PFOS interaction at the metabolome level, identifying MPs as dynamic modulators of PFOS uptake. Co-exposure elicits an emergent stress signature involving redox imbalance and a tissue-specific secondary metabolism allocation. Our data should be carefully considered concerning environmental PFOS contamination in the context of environmental microplastics.

Our study showed a moderate knowledge and practice regarding breast cancer screening, alongside a generally good attitude. The factors associated with good practice underscore the importance of educational interventions and the targeted screening strategies.