Wireless network performance depends extensively on the carrier interference. Inter-carrier interference (ICI) can be caused by several factors, including carrier frequency offsets, Doppler spread due to channel time variation, and sampling frequency offsets, which degrade the performance of an orthogonal frequency division multiplexing (OFDM) system. Hence, reducing ICI is a major task in communication systems. The lower the ICI, the higher the performance of the OFDM system. Moreover, the application of deep learning has demonstrated significant improvements in communication reliability and reduced the computational complexity of 5G and subsequent networks. Deep learning combined with ICI self-cancellation techniques yields promising results. In this paper, a hybrid framework leveraging mirror-based techniques is developed to enhance data robustness and mitigate interference while using deep learning models for adaptive, real-time ICI prediction and suppression. We incorporated multiple symbol rates and multi-carrier vector transmission at the physical layer to improve signal resilience against channel imperfections. These techniques were used in conjunction with deep learning models such as long short-term memory (LSTM) to predict ICI patterns and dynamically adjust mirror-based symbol-repetition parameters to optimize signal quality. The LSTM was incorporated with an attention mechanism. Improved interference cancellation was observed through the combined strength of symbol repetition and adaptive neural network-based interference prediction. The performance of the proposed work was evaluated using two modulation techniques: quadrature phase-shift keying and binary phase-shift keying. The results for bit error rate and carrier-to-interference ratio were best with ICI self-cancellation combined with LTSM. In summary, the proposed approach improves OFDM system performance by eliminating ICI and enhancing signal quality.

Systematic innovation methods resolve technical contradictions effectively but lack procedures for cultural–technical contradictions, in which scientifically accurate information conflicts with cultural taboos. Such contradictions pervade sensitive domains, including sexuality education, mental health, and death education, where feasibility depends on both scientific integrity and social acceptance. Theory of Inventive Problem-Solving (TRIZ) addresses physical contradictions but cannot operationalize cultural constraints, while participatory design lacks algorithmic protocols for conflicting stakeholder standards. This study develops the culturally constrained sensitive content design (CSCD) method, a five-module framework with formalized decision protocols that operationalizes cultural adaptation through tiered content architecture, metaphor-based narratives, de-realization techniques, and multi-stakeholder validation. The framework was validated through interviews with 20 children and a 7-expert focus group, using sex education comics in southern China as an extreme-constraint case. The CSCD method offers the first algorithmic framework for cultural–technical contradictions, with structural analogies suggesting transferability to mental health, death, and financial literacy education pending future empirical validation. It provides replicable procedures for materials that require both scientific accuracy and cultural acceptance, operationalizes cultural adaptation into designable parameters, and indicates potential transferability to other sensitive domains requiring multi-stakeholder validation.

The rapid diffusion of artificial intelligence (AI), particularly generative AI tools, has significantly reshaped educational assessment practices, creating new opportunities and challenges for feedback, academic integrity, and curriculum design. Despite the growing volume of scholarship, research in this area remains fragmented, making it difficult to discern dominant trends, key contributors, and emerging themes. This study employs a bibliometric review to map the intellectual structure and evolution of research on AI in educational assessment between 2015 and 2025. Using open-access journal articles indexed in the Dimensions.ai database and aligned with Sustainable Development Goal 4 (Quality Education), a curated dataset of 89 studies was analysed. Bibliometric techniques, including co-authorship, citation, co-citation, and keyword co-occurrence analyses, were applied using VOSviewer and descriptive statistics. The findings reveal a sharp growth in publications following the emergence of generative AI, with influential clusters focusing on assessment redesign, feedback, academic integrity, and higher education applications. While established institutions and authors dominate the field, collaboration networks remain fragmented, and contributions from non-Western contexts are comparatively limited. The study highlights the need for stronger international collaboration and context-sensitive research to support equitable and responsible integration of AI in educational assessment.

Lung cancer is a leading cause of cancer-related mortality due to delayed diagnosis and limitations of conventional screening methods. This study presents an artificial neural network-based computer-aided diagnosis system for automated lung cancer detection using computed tomography scans. The proposed methodology integrates advanced digital image processing techniques with machine learning-based classification to improve diagnostic accuracy and reliability. The framework consists of multiple stages. Initially, computed tomography images are preprocessed using a two-dimensional median filter to suppress noise while preserving structural boundaries. Morphological operations and contrast enhancement techniques are applied, followed by adaptive thresholding to segment lung regions. A seeded region growing technique is then employed to identify suspicious regions and extract relevant image segments. From these segments, 25 texture features are computed using the gray level co-occurrence matrix, capturing statistical properties such as contrast, correlation, entropy, homogeneity, and energy. Two artificial neural network classifiers, namely the back propagation neural network and the radial basis function neural network, are trained using these features. A dataset of 500 computed tomography images, including both cancerous and non-cancerous cases, is used for performance evaluation. Experimental results demonstrate that both models achieve high classification accuracy, sensitivity, and specificity, while the radial basis function neural network consistently outperforms the back propagation neural network, achieving a maximum accuracy of 94%. These findings highlight the effectiveness of the proposed system as a reliable, non-invasive, and computationally efficient tool for early lung cancer detection, with potential applicability to other medical imaging domains.

Conventional two-dimensional integrated circuits are increasingly limited by interconnect delays, power density, and scaling constraints predicted by Moore’s law. Three-dimensional (3D) integrated circuits (ICs), enabled by through-silicon via (TSV) technology, overcome these limitations by vertically stacking dies, thereby reducing interconnect lengths, increasing bandwidth, enhancing functionality, and allowing higher integration density. However, noise coupling in TSV-based 3D ICs significantly impacts signal integrity, especially at high operating frequencies. This study proposes replacing the traditional dielectric silicon dioxide (SiO₂) with Teflon due to its lower dielectric constant and higher thermal resistivity. It provides a comprehensive comparative analysis of copper (Cu), carbon nanotube (CNT), and conventional semiconductor core materials using both single-liner and stacked-liner configurations with SiO₂ and Teflon dielectrics at 10 GHz and 1 THz. Noise coupling is assessed in terms of electric potential and expressed as attenuation in dB. Results show that CNT interconnects consistently display lower noise coupling than metallic and semiconductor cores. At 10 GHz and 4 µm arc length, Teflon–CNT achieves 10.75 dB compared to 5.03 dB for SiO–Cu. At 1 THz, attenuation decreases to 13.56 dB, representing an 8.53 dB reduction relative to Cu. Although SiO2 remains an industry-standard dielectric, the Teflon-based stacked configuration offers superior high-frequency isolation. Consequently, the CNT–Teflon structure emerges as a promising solution for next-generation high-performance 3D IC systems beyond conventional scaling limits.

In recent years, artificial Intelligence (AI) has emerged as one of the main global trends in innovation and competition. Businesses have begun using AI to boost productivity, reduce costs, and improve strategic decisions. This research examines how AI affects internal organizational processes (IOP) by examining the food wholesalers industry in the United Arab Emirates (UAE). A quantitative research design was used, and the data were collected via an online survey administered via Google Forms, with 218 wholesalers in the UAE participating. SPSS 26 was used to analyze the data using descriptive statistics, correlation analysis, and multiple regression. The findings indicated that AI had a significant effect on the IOP. The greatest impact was demonstrated by AI infrastructure, followed by employee readiness and leadership support. The research concludes that enhancing IOP requires a combined strategy that includes technological preparedness, the role of leaders, data quality, and staff competence. It provides recommendations, limitations, and future research directions to guide the organization and researchers in their future research on the drivers of internal process improvement.