Modern University for Business and Science

Machine learning has achieved notable success in medical imaging; however, its reliability remains constrained by the quality and quantity of training data, particularly in specialized domains such as fetal brain ultrasound. Federated learning (FL) offers a collaborative alternative by enabling multiple institutions to train shared models without exposing sensitive data. Yet, its effectiveness often degrades under non-independent and identically distributed (non-IID) data and its communication demands can be prohibitive. To address these challenges, we propose FedFB, a communication-efficient federated learning framework that integrates online ensemble knowledge distillation for privacy-aware collaborative learning. In FedFB, multiple teacher models trained locally on private datasets collectively distill their knowledge into a lightweight dual-branch student network enhanced with an auxiliary attention–convolution module, without increasing model complexity. The distillation process leverages a small auxiliary public dataset for soft-label transfer, reducing data exposure by avoiding the exchange of raw private data or model parameters, while relying on shared supervisory signals for knowledge transfer. Experimental results on fetal brain ultrasound datasets demonstrate that FedFB achieves 91.07% accuracy, 91.28% precision, 91.03% recall, and 91.15% F1-score under the non-IID conditions, while reducing communication overhead by more than 85% compared to traditional FL methods. Furthermore, the framework’s robustness was validated on brain tumor MRI and chest X-ray datasets, confirming its generalization capability across distinct medical imaging modalities.

Dextrose monohydrate-capped Ni 0.6 Zn 0.2 Sb 0.2 Fe 2 O 4 offers tunable optical band gaps and high methyl violet adsorption, showing significant potential for synergistic multifunctional optoelectronic and environmental applications.

Lebanon’s ongoing financial, banking, and political crises disrupted education, exposing the often overlooked influence of neighboring activity systems. This case study investigated remote learning at a private university in Lebanon through Cultural-Historical Activity Theory (CHAT), aiming to fill a gap in cultural-historical research on remote learning and further develop CHAT. Following an abductive approach, data was collected through in-depth interviews with 7 university students. Guided by Engeström and Sannino's frameworks, the study explored contradictions students faced during remote learning and how they addressed them. Students reported challenges including internet issues, demotivation, and difficulty concentrating. Most contradictions (61 out of 79) remained unresolved. When resolved, contradictions were mainly addressed by modifying tools used in remote learning. Students who modified rules of interaction or engaged more actively with the learning community were able to overcome further challenges. The study also underscored the importance of “physical space” in remote education; it proposes a 7th component to CHAT, the Extended Space (E.S), which incorporates physical and virtual environments where learning occurs. Ultimately, this study contributes to the development of CHAT by highlighting the spatial dimensions and interactivity of remote learning contexts, while also providing practical recommendations for optimizing learning in digital environments.

BackgroundThe Mediterranean diet (MeD) is associated with favorable pregnancy outcomes, but contribution of this dietary pattern during pregnancy with small (SGA), appropriate (AGA) and large (LGA) for gestational age births is limited.MethodsFor this prospective national cohort study 618 Lebanese pregnant women were recruited. Infant birth weight was categorized into SGA (n = 73), AGA (n = 447) and LGA (n = 98). Modifiers of birth weight outcomes included dietary adherence to Lebanese MeD (LMeD), trimestral and total weight gain, MAP (mean arterial pressure) and PP (pulse pressure) and psychosocial, socio-demographic, and maternal health factors. Descriptive statistics compared differences among SGA, AGA, and LGA infants. Hierarchical linear regression modeling identified determinants for birth weight categories and hierarchical logistic regression modeling was used to identify factors associated with increasing the likelihood of SGA or LGA compared to AGA births.ResultsAdherence to the LMeD was associated with AGA birth weights where intakes of dairy products were associated with lower normal AGA births and dried fruits with higher normal AGA births. Adherence to LMeD did not enter models for SGA or LGA but intakes of specific foods and maternal health status indicators did. For SGA infants, appropriate gestational weight gain (GWG) mitigated against a low birth weight whereas higher burghul intake in T1 and higher MAP in T2 and T3 were linked to increased SGA risk. For LGA infants, greater parity, previous macrosomia and poor sleep quality in T3, and higher intake of olive oil in T2 were associated with higher risk of a LGA birth whereas higher PP in T1 decreased the odds of a LGA birth.ConclusionsScreening for family history of diabetes and macrosomia, targeting trimester-specific gestational weight gain, monitoring maternal blood pressure, pulse pressure, and sleep quality, and promoting adherence to the Lebanese Mediterranean diet are important strategies to optimize infant birth outcomes.