Spectrum of Engineering Sciences

A DEEP LEARNING APPROACH TO PCOS DIAGNOSIS: TWO-STREAM CNN WITH TRANSFORMER ATTENTION MECHANISM

Wed, 02 Jul 2025 00:00:00 +0300

Polycystic Ovary Syndrome (PCOS) is a prevalent endocrine disorder affecting a significant portion of women worldwide, often underdiagnosed due to the complexity of its symptoms and limitations in existing diagnostic tools. With advancements in deep learning and medical imaging, automated classification systems offer the potential to revolutionize PCOS detection through precision and scalability. This study proposes a novel Two-Stream Convolutional Neural Network (CNN) architecture enhanced with Transformer-based attention mechanisms for classifying PCOS from ultrasound images. Leveraging dataset of 11,784 images, our framework splits each ultrasound image into upper and lower halves to capture anatomical variance and apply convolutional encoding separately. A Multi-Head Attention layer then integrates spatial dependencies between the two streams, enhancing feature discrimination and improving model interpretability. Experimental evaluations show that the proposed model achieves a classification accuracy of 98.96%, an F1-score of 0.99, and minimal loss on the test dataset. These results highlight the model’s robustness and potential applicability in real-world clinical settings for the early detection of PCOS.

HYBRID DEEP LEARNING MODELS FOR MULTI-CLASS CLASSIFICATION OF CHEST X-RAY IMAGES: NORMAL, PNEUMONIA, AND COVID-19

Ayesha Saddique, Abdul Manan, Muazzam Ali, Sidra Siddiqui, Muhammad Rehan — Wed, 02 Jul 2025 00:00:00 +0300

CXRs remain a major diagnostic method that assists in the identification of respiratory diseases such as COVID-19, pneumonia, and tuberculosis. They are helpful in the clinical setting, especially in low and middle-income countries, where they serve as the gateway to the healthcare system as a result of their affordability when compared to CT and MRI scans. In spite of these advantages, CXR scans still exhibit considerable challenges, particularly in diagnosing CXRs which remains a labor-intensive high expertise process with a large range of inter-reader variability. The problem is exacerbated by multi-class classification where there is pneumonia and COVID-19 which have overlapping radiographic features. The problem that this particular work intends to address is developing and testing the hybrid models that consists of CNNs and transformers models to increase diagnostic accuracy for classification of chest X-ray images into Normal, Pneumonia, and COVID-19 categories. The dataset used was comprised of 7,135 chest X-ray images, which after were subjected to uniform pre-processing to aid in consistency and standardization. Hybrid models were developed such that they paired CNNs with other transformer-based models like DenseNet121 + Swin Transformer and EfficientNetB0 + Vision Transformer. With these models, training was undertaken using the TensorFlow/Keras framework and evaluation was done based on accuracy, precision, recall, F1 score, and confusion matrix. The findings indicate that the DenseNet121 + Swin Transformer model achieved the highest accuracy, precision, and recall scores, outperforming all other models, which demonstrates its potential for more reliable classification compared to CNN-based techniques. The study nonetheless notes the considerable potential of such hybrid models to augment diagnostic functionalities in clinical practice, even with hurdles like dataset imbalance and the absence of real-world validation.

AI-BASED RESOURCE MANAGEMENT FRAMEWORK FOR NEXT-GENERATION WIRELESS NETWORKS

Rahat Ullah, Shafiq Ur Rahman, Zubair Khalid, Sabeen Asghar, Hidayat Ullah — Wed, 02 Jul 2025 00:00:00 +0300

The advent of sixth generation (6G) wireless networks brings unprecedented challenges in managing ultra-dense, dynamic, and heterogeneous environments. Classical optimization methods lack the scalability, adaptability, and selflessness required to address the challenge of resource management. This paper proposes an AI-enabled resource allocation framework specifically designed for 6G networks through the integration of state-of-the-art machine learning (ML), deep reinforcement learning (DRL), federated learning (FL), and graph neural networks (GNNs). The envisioned multi-layered architecture allows edge devices, UAVs, and base stations to perceive nearby environments, forecast traffic trends, make in-real-time decisions, and jointly train models with privacy preserved. An end-to-end global controller from GNN provides orchestration over the network topology. We review state-of-the-art AI methods and discuss their adequacy in accommodating resource allocation complexity with trade-offs between convergence, latency, and scalability. We conclude by describing current challenges—heterogeneous data, stable convergence, and limited computations—and sketch future directions of research towards reliable, explainable, and energy-efficient AI deployment in 6G systems.