Imagine having a digital copy of yourself that doctors can use to test different medicines before you actually take them. Usually, this requires sharing private medical records in one big pile, which is a huge privacy risk. This project creates a way for AI to learn from 800,000 patients across the globe without the data ever leaving its original home.
Healthcare providers lack guidelines to predict how a specific type 2 diabetes patient will respond to a specific treatment. Current big data efforts are blocked by strict privacy laws like GDPR.
A federated database network (dAIbetes-Net) and a software platform that hosts AI virtual twin models for personalized treatment prediction.
If you are a health-tech software provider dealing with strict GDPR privacy laws — this project developed a federated health data platform that allows AI training without uploading sensitive patient data to a common cloud. This enables the creation of virtual twin models for personalized glycemic control.
If you are a drug development company dealing with the lack of guidelines for individual treatment outcomes — this project developed prognostic virtual twin models. These models aim to reduce prediction errors by at least 10% compared to population average models.
If you are a private clinic network dealing with the high cost of diabetes management, which costs roughly 893 billion EUR annually — this project developed a software platform for clinical application of virtual twins. This allows for personalized prediction of treatment outcomes for specific patients.
Based on available project data, no commercial pricing is listed; however, the project is supported by an EU contribution of EUR 8,938,145.
Yes, the project is designed for scale by harmonizing data from approximately 800,000 patients across 6 global cohorts using a federated database network called dAIbetes-Net.
Based on available project data, specific licensing terms are not provided, but the project involves a consortium of 13 partners including 3 industry members and 3 SMEs.
It uses federated learning, which allows AI to learn from data without requiring the upload of sensitive patient information to a common cloud, specifically addressing GDPR constraints.
The project period runs from 2024-01-01 to 2028-12-31, indicating a multi-year development and validation phase.
The consortium is well-balanced for commercialization, featuring 13 partners across 7 countries. With a 23% industry ratio (including 3 SMEs), there is a clear bridge between the 6 universities and 2 research institutes and the actual market. The inclusion of US partners alongside EU members suggests a global strategy for data acquisition and market entry.
University of Hamburg
Contact us to explore licensing opportunities for federated AI health platforms.
