Author(s)

Chenjue Yu

Corresponding Author

Chenjue Yu

Abstract

Breast cancer mortality disparities between rural and urban populations highlight systemic inequities in healthcare access, particularly in low-resource settings. To address this challenge, we developed an AI-enhanced mobile screening framework integrating portable ultrasound technology, convolutional neural networks (CNNs), and federated learning to improve early breast cancer detection in rural China. Our approach combines mobile diagnostic units with an AI model trained on a multi-center dataset of 128,432 mammographic and ultrasound images, validated against pathological gold standards. The system employs a two-stage workflow: (1) automated lesion segmentation using a U-Net-based architecture, and (2) malignancy classification via a hybrid CNN-radiomics model, augmented by generative adversarial networks (GANs) to address class imbalance. Deployed across five provinces, the mobile network achieved a 92.3% diagnostic accuracy (sensitivity: 90.7%, specificity: 93.5%) in field trials, surpassing primary care physicians performance by 18.7%. Implementation data revealed a 3.2-fold increase in screening coverage (63.4% vs. 19.7%) and a 17.3% rise in early-stage cancer detection rates within six months. Cost analysis demonstrated a 64% reduction in per-screening expenses compared to conventional methods, with AI pre-screening reducing radiologists’ workload by 78%. Notably, the edge-computing architecture enabled offline operation in areas with limited internet connectivity, ensuring scalability. While challenges persist in model generalizability across diverse ethnic populations and long-term follow-up integration, this study demonstrates that AI-augmented mobile screening can significantly mitigate rural-urban disparities in cancer care. By decentralizing diagnostic capabilities and enhancing resource efficiency, our framework offers a replicable model for global rural health systems aiming to achieve equitable cancer outcomes.

Keywords

Breast Cancer Screening, Artificial Intelligence (AI), Rural Healthcare Disparities, Early Cancer Detection