Current Status of Machine Learning and Artificial Intelligence in Cervical Cancer Screening and Diagnosis: A Systematic Review

Abstract

Background: Cervical cancer poses a substantial global health challenge, predominantly affecting underprivileged countries. The limitations of current screening methods, such as cytology and visual examination, underscore the need for improved techniques. Artificial intelligence (AI) and machine learning (ML), particularly convolutional neural networks, offer promising solutions in this regard. Methodology: Fifteen studies meeting the inclusion criteria were examined. The PRISMA criteria guided the exploration of cervical cancer screening studies employing AI, ML, and deep learning on PubMed/MEDLINE and Google Scholar. The search focused on "artificial intelligence" and "Pap smear." The investigation specifically delves into English-language studies post-2019 that pertain to the machine learning and deep learning classification of cervical cancer using mobile devices. Histology, animal research, and pre-2019 investigations are excluded. Titles and abstracts were carefully reviewed for any discrepancies and subsequently discussed. The process of data extraction involved compiling information from the selected articles. Result: The systematic review investigates the impact of AI and ML on cervical cancer detection, screening, and diagnosis. Our review reveals enhanced accuracy and efficiency in innovative technologies such as CytoBrain and computer-aided diagnostic systems employing Compact VGG and ResNet101. ML techniques like logistic regression, MLP, SVM, KNN, and naive Bayes prove beneficial for managing complex datasets, particularly when combined with class-balancing procedures. The promising aspects include the application of deep learning for automation and AI-assisted digital microscopy. These findings signify a transformative shift in cervical cancer screening, underscoring the potential of ML and AI technologies to elevate diagnostic accuracy and accessibility. Conclusion: Our study demonstrates advancements in both accuracy and responsiveness. Despite recognizing scientific and ethical considerations, the study underscores the potential of AI to enhance cervical cancer care. This systematic review advocates for policymakers and healthcare practitioners to use ongoing research for informed decision-making in this rapidly evolving field.