-I am positively surprised that we were selected, cause the eye of the needle was narrow. The list of applicants was full of heavy players, says general manager Karl Trygve Kalleberg in Age Labs.
Detected too late
Most age-related diseases are only discovered when symptoms occur. Then it is difficult, if not impossible, to reverse the damage. This applies, for example, to cardiovascular disease, cancer and dementia. There is a growing belief and focus that personalized and preventive medicine will help us to postpone or avoid disease, by taking action long before the disease manifests itself.
"Algorithm-driven blood tests"
Age Labs core product is a platform for "algorithm-driven blood tests". By reading the epigenetic profile of a patient from a blood test, they can use algorithms to identify patterns that say something about future disease development. However, this type of diagnostics depends on extensive amounts of health data during the development process, and has a great potential for continuous improvement after it is installed in daily operation, in the form of learning the algorithms.
The training of machine learning algorithms depends on large amounts of sensitive health data, both genetic and epigenetic data in addition to information about diseases and disease development. The privacy challenges in the project, which they want to explore in the sandbox, are how to ensure anonymity of the data they have collected, so that they can be shared freely? And how should they proceed, in terms of anonymization and data minimization, when they are going to learn the algorithms after they went into production?
The project plan for the sandbox work has been laid out. In it, Age Labs and the Norwegian Data Protection Authority have focused on the issues they will be working on in the sandbox. Age Labs is in the FoU stage of its first diagnostic test for rheumatological disease, and in the sandbox project they will focus on the possibility of anonymizing, possibly pseudonymizing, health data that is necessary for this particular test. Specific focus; how to ensure anonymization of epigenetic data?
In addition, the project will explore an important issue of principle in the KI context. With current legislation, there is a somewhat freer rein to the use of personal data, if it happens in connection with health research and development. But as soon as a product as a result of the research is put into production and comes into commercial use, the exemption from the general privacy legislation lapses. With many AI products, there is great potential for post-learning, ie continuous improvement of the product based on data collected during use. The sandbox will look at how Age Labs can proceed, with a view to anonymisation and data minimization, when they will learn the algorithm after it has been put into production, and which legal basis / treatment basis will then be used.