Federated XGBoost-Powered Web Application for Safeguarding Personal Health Data in Hospitals and Clinics of Metropolitan Lima

Abstract

In the current context of digitalization, artificial intelligence (AI) has become a key opportunity for transforming operations in the healthcare sector. However, its adoption faces challenges related to data privacy, trust in emerging technologies, and technical limitations, particularly in medical institutions in developing countries. Although several studies have explored federated learning (FL) for preserving privacy in clinical environments, there remains a gap in its application within contexts that face limited infrastructure and operate under specific regulatory frameworks, such as in Peru. In this study, we present a federated-learning-based web platform integrating an XGBoost model to diagnose premature births without centralizing sensitive clinical data. The system was designed to operate across multiple healthcare institutions in Metropolitan Lima, ensuring regulatory compliance and preserving data privacy. Preliminary results demonstrate the technical feasibility of this approach, highlighting an effective balance between privacy, diagnostic accuracy, and adaptability to limited infrastructures. This proposal contributes to the secure adoption of AI in Latin America's healthcare sector.

Ähnliche Arbeiten