[Summer Rewind] Federated Learning: Training AI Models Without Sending Data

Did you miss these insights published this year? Now is the time to catch up. This summer, we offer you the opportunity to discover or rediscover the medical news published in our affiliated publication, DirectIndustry e-magazine. In this article, we focus on data governance in healthcare and the power of federated learning to train AI models without sending data.

While artificial intelligence holds promise for remarkable technological advancements in healthcare, limited access to data remains a significant hurdle in Europe. Federated Learning emerges as a solution to maintain data confidentiality while training high-quality AI models.

Here’s how Federated Learning works:

• Instead of consolidating data on a single central server, federated learning allows data to remain distributed across independent centers.
• Each center keeps its data secure, while algorithms and predictive models move between them.
• Researchers and companies are increasingly adopting this approach to address data governance issues and escape biases from single-centric studies.

Notable initiatives include:

• The CEA-List’s project combines federated learning and blockchain to create a collaborative platform for training AI models without sharing sensitive information.
• French biotechnology company Okwin’s AI models accurately predict mesothelioma patients’ response to neoadjuvant chemotherapy using interpretable data from four French hospitals.
• The French National Institute for Research in Digital Science and Technology (INRIA) also contributes to this innovative AI model, fostering a more robust and ethically aligned AI landscape in healthcare.

Explore the full article for more info: Data Governance in Healthcare: The Power of Federated Learning or How to Train AI Models Without Sending Data