Machine Vision Approaches for Cervical Cancer Screening Using Pap-Smear Images: A Systematic Review.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Cervical cancer remains a major cause of mortality, particularly in developing countries.
• The Pap smear test is crucial for early detection, with machine learning enhancing diagnostic accuracy.
• Machine vision frameworks are increasingly applied to Pap smear image analysis for improved screening.

Purpose of the Study:

• To systematically review machine vision-based frameworks for cervical cancer screening using Pap smear images.
• To analyze segmentation, feature extraction, and classification methods in current machine learning approaches.
• To evaluate the validity and relevance of commonly used datasets for cervical cancer screening.

Main Methods:

• Systematic review following PRISMA guidelines.
• Analysis of machine vision techniques including segmentation, feature extraction, and classification.
• Examination of datasets used in machine learning for Pap smear analysis.

Main Results:

• Deep learning methods achieve higher accuracy with large, annotated datasets.
• Lack of high-quality, multi-cell datasets and inconsistent evaluation metrics impede direct comparison of studies.
• No single framework consistently outperforms others across all scenarios.

Conclusions:

• Significant progress has been made, but challenges remain in generalization and clinical implementation.
• Deep learning and hybrid models show potential but require robust datasets and standardized evaluation.
• Future research should focus on creating standardized datasets and evaluation frameworks for improved clinical applicability and system scalability.