Governance-Aware Privacy-Preserving AI Infrastructure for Ghana’s Maternal Health Ecosystem

James Tetteh Matey

Keywords: Federated Learning, Differential Privacy, Data Governance, Maternal Health, AI Infrastructure

TL;DR: This study proposes a governance-aware privacy-preserving AI infrastructure using federated learning and differential privacy for maternal health risk prediction.

Abstract: The Maternal health data in Ghana is sensitive and difficult to share because of privacy and governance concerns. This study proposes a governance-aware privacy-preserving AI infrastructure using federated learning and differential privacy for maternal health risk prediction. A neural network model was trained using centralized learning, federated learning, and federated learning with Gaussian noise. The centralized model achieved 68.97% accuracy. The federated model improved to 70.94% after 20 rounds. When differential privacy was added, the accuracy was 68.47%, showing only a small performance drop. High-risk cases were still well detected under privacy protection. These findings are consistent with existing healthcare federated learning studies. This research connects data governance, privacy, AI, and healthcare data infrastructure, and shows how Ghana can enable collaborative health analytics without sharing raw patient data.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 1