Recent Advances in Automatic Cervical Cancer Detection and Segmentation Using Sparsity-Aware Orthogonal Initialization in Deep Neural Network Classifiers: A Systematic Review

Abstract

Cervical cancer remains a major cause of cancer-related mortality among women worldwide, particularly in low- and middle-income countries where access to early diagnostic facilities is limited. Early detection through screening methods such as Pap smear, HPV testing, and colposcopy is crucial for reducing mortality rates; however, manual examination of cervical cell images is often time-consuming, subjective, and prone to human error. In recent years, deep learning-based approaches have significantly enhanced the automation of cervical cancer detection and segmentation tasks. Convolutional neural networks (CNNs) have demonstrated strong performance in identifying abnormal cellular patterns, while advanced segmentation models such as U-Net and nnU-Net enable precise delineation of tumor regions with high accuracy. A notable advancement in this area is sparsity-aware orthogonal initialization (SAOI), which improves training efficiency and convergence by enabling sparse yet effective neural architectures, thereby reducing computational complexity without compromising performance. This review examines recent advancements in deep learning architectures, segmentation techniques, and optimization strategies, highlighting key trends, challenges, and future directions such as explainable AI, multimodal data integration, and improved clinical validation for real-world applicability.