x
Loading
+ -

Personalized medicine with optimum data protection

open hand presenting a network of dots and lines
The Clinnova research consortium relies on decentralized AI training to enable personalized therapies for autoimmune diseases. (Image: AdobeStock)

Researchers from Switzerland, France, Germany and Luxembourg plan to use artificial intelligence to improve the treatment of patients with autoimmune diseases. The aim is to enable customized therapies for multiple sclerosis, rheumatoid arthritis and inflammatory bowel conditions.

15 May 2024 | Angelika Jacobs

open hand presenting a network of dots and lines
The Clinnova research consortium relies on decentralized AI training to enable personalized therapies for autoimmune diseases. (Image: AdobeStock)

The same treatment doesn’t work the same in every patient. Every person is different, has their own individual disease progression, or responds better or worse to specific active ingredients – with more or fewer side effects. It is for this reason that the concept of “big data” is raising hopes in the healthcare sector. The analysis of large quantities of patient data could yield new insights into who will benefit the most from which therapy.

One major obstacle to this is that the data is so heterogeneous: hospitals sometimes have their own procedures that define how they collect and analyze samples or which measurements they take. This makes it difficult to compare data from different health centers.

Another problem is that in order to bring together sufficient quantities of data that could be used to train an artificial intelligence (AI), it is necessary to combine patient data across borders. This raises concerns around data protection.

Decentralized AI training

The research consortium Clinnova wants to realize the full potential of AI for precision medicine. With this in mind, the project approach is intended to ensure the quality of the data as well as data protection: on the one hand, the participating institutions establish joint processes in order to collect a series of clearly defined data over the coming years. This includes clinical data, biological samples such as blood and urine, and measurements with digital sensors, as well as medical imaging data. To this end, the researchers are recruiting patients in order to establish several cohorts, which they will accompany over the next few years.

On the other hand, they are developing digital infrastructure for “federated learning.” The principle behind this decentralized training for AI is that the data from different health centers does not flow into a large database in order to train the AI; rather, parts of the algorithm are applied to the respective data of the individual institutions. The data remains decentralized, and the AI is trained using aggregated statistical parameters derived from the data.

To top