“Artificial Intelligence-Based Early Detection of Cervical Cancer Using Pap Smear Images: A Comprehensive Systematic Review”

Abstract

Background: Cervical cancer remains a major public health concern, particularly in regions where access to regular screening and expert cytopathological assessment is limited. Pap smear cytology is an established screening method; however, manual examination of cervical cell images can be time-consuming and is subject to inter-observer variability. Recent advances in artificial intelligence (AI), machine learning (ML), and deep learning (DL) have demonstrated significant potential for improving the automated detection and classification of abnormal cervical cells. Objective: This review aims to comprehensively examine the application of AI-based techniques for the early detection and classification of cervical cancer and precancerous lesions using Pap smear images. It focuses on commonly used image-processing approaches, machine learning and deep learning architectures, publicly available datasets, classification strategies, and reported diagnostic performance. Methods: Relevant studies investigating AI, machine learning, deep learning, and computer vision techniques for cervical cytology image analysis are reviewed. The literature is examined with emphasis on image preprocessing, segmentation, feature extraction, classification, convolutional neural networks (CNNs), transfer learning, and performance evaluation using measures such as accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). The methodological characteristics, datasets, and limitations of the selected studies are also critically assessed. Results: Existing research indicates that AI-based approaches can assist in distinguishing normal and abnormal cervical cells and may improve the consistency and efficiency of Pap smear image interpretation. Deep learning and transfer-learning methods have shown promising results in automated feature extraction and classification. However, differences in datasets, image quality, preprocessing methods, classification categories, evaluation protocols, and external validation make direct comparison between studies challenging. Limited dataset diversity and insufficient validation on real-world clinical data remain important barriers to clinical implementation. Conclusion: AI-assisted analysis of Pap smear images represents a promising approach for supporting early cervical cancer detection and improving cervical screening workflows. Future research should focus on large and diverse datasets, standardized evaluation protocols, explainable AI, external clinical validation, and integration of AI systems into routine screening environments. Robust prospective studies and clinically validated models are essential before widespread deployment in healthcare settings.