A short review of technologies and AI/machine learning for optimal control of human papillomavirus

Human papillomavirus (HPV) control has shifted from a narrow screening problem to a multiscale decision problem involving vaccination, triage, population outreach, risk stratification, follow-up, and treatment referral. Literature reviews show three dominant technological directions: (i) mathematical and decision models for optimizing HPV prevention strategies, (ii) digital and molecular technologies that expand screening access, and (iii) artificial intelligence (AI) and machine learning (ML) tools for image analysis, cytology, and individualized risk prediction. Recent studies indicate that self-sampling, HPV genotyping, molecular triage, automated visual evaluation, AI-assisted cytology, and data-driven risk models can improve coverage and targeting, especially in low-resource settings. At the same time, the literature suggests that true closed-loop optimal control of HPV remains underdeveloped: most studies optimize policies through simulation, cost-effectiveness analysis, or supervised prediction, while only a small number explicitly formulate and solve control problems with reinforcement learning. Overall, the evidence supports a hybrid future in which HPV control is organized as an integrated system: vaccination and screening policies are optimized using mathematical models, operational delivery is strengthened by digital technologies, and clinical triage is refined using AI/ML. For practical deployment, such systems must also represent clinical, behavioral, implementation, and measurement disturbances and combine anticipatory feedforward action with outcome-based feedback whenever disturbances can be monitored. The main research gap is the lack of prospective, real-world validation of these components as a unified decision architecture.