Federated Learning in Digital Health: Addressing Privacy Concerns and Advancing AI Solutions

Artificial Intelligence (AI) has the potential to revolutionize the landscape of Digital Health, offering unprecedented opportunities for enhanced diagnosis, patient monitoring, and treatment planning. The vast troves of data generated across hospitals worldwide are the main driving force for creating sophisticated AI-driven solutions for digital health. However, privacy concerns have long been a major barrier to data sharing in healthcare because patient data is highly sensitive, and regulations such as HIPAA in the United States and GDPR in Europe impose strict requirements for its protection. Federated Learning? (FL) offers a promising solution to this privacy issue, thereby paving the way for unprecedented advancements in healthcare. FL allows multiple hospitals to collaboratively learn AI models without compromising data privacy. During model learning, the hospitals only share the knowledge they have learned from their data instead of raw data with each other. This decentralized approach ensures that sensitive patient information remains secure and confidential, thereby addressing one of the most pressing obstacles in Digital Health. However, implementing FL requires careful coordination among participating hospitals and robust security measures to protect against potential attacks.

Additionally, ensuring the quality and consistency of data across disparate sources remains a concern. Our researcher Faisal Ahmed , under the continuous supervision of the CRISES group at Universitat Rovira i Virgili , focuses on addressing these challenges in his PhD thesis. He will validate his thesis by developing AI models for breast cancer relapse and molecular subtype prediction, analyzing radiological and histopathological images from hospitals associated with the EU-funded BosomShield project, where NVISION is one of the partners.?