The Role of Artificial Intelligence in Cervical Cancer Screening: From Pap Smears to Deep Learning

Cervical cancer screening has evolved from conventional Pap smears to liquid‐based cytology and HPV DNA testing, and now increasingly leverages AI-driven image analysis. Modern AI approaches – especially convolutional neural networks (CNNs) – can automatically detect and classify cells in digital cytology and whole‐slide images. Emerging transformer‐based models further integrate multimodal data (HPV status, cytology, colposcopy images) to enhance diagnostic sensitivity and reduce missed cases. These tools markedly increase detection rates while alleviating the burden on human screeners. In practice, portable AI-assisted devices are being fielded: for example, smartphone colposcopes with embedded deep learning have demonstrated CIN2+ sensitivities (>90%) that exceed expert readers. Cloud-linked systems (e.g. MobileODT’s EVA Scope) enable centralized analytics and regular model updates, while on-device AI can operate offline for low-connectivity settings. However, limitations remain. Algorithmic bias from non-representative training data can skew performance, and the “black box” nature of deep models raises interpretability and trust issues. Regulatory and privacy gaps must be addressed – clear guidelines are needed to ensure patient data security and to define AI’s role as an adjunct (not a replacement) to clinician judgment. Moving forward, robust validation in diverse populations, human–AI collaboration in screening workflows, and equitable deployment in underserved regions are essential. By coupling AI with human expertise and broad access, these innovations can help reach WHO targets (70% screening by 2030) and accelerate progress toward global cervical cancer elimination