Healthcare systems are changing. Their transformation is driven by a widely acknowledged fact: no single institution has the data required to push forward research, improve healthcare pathways, and develop clinically useful and responsible AI models, at least not alone.<\/p>\n
This conclusion became clear to the Swiss University Hospitals in 2016. They realized that to deliver on the promise of personalized medicine, medical teams need more data\u2014more relevant data, more comparable data, more representative data.<\/p>\n
As a result, the Swiss Personalized Health Network<\/em> embarked on an ambitious project with the goal of allowing healthcare facilities to work together to analyze their data without ever losing ownership of it. A formidable scientific, technical, regulatory and logistical challenge.<\/p>\n Fr\u00e9d\u00e9ric Pont, cofounder and COO of Tune Insight, explains how this moment marked the start of the adventure that would cumulate in the founding of Tune Insight.<\/p>\n \u201cTo be truly personalized, clinical models have to be able to compare a specific patient to other individuals with similar characteristics. But the more precise these criteria become, the more unlikely it is that a single hospital will have enough similar cases. So expanding the research to other facilities becomes essential<\/em>.\u201d explains Fr\u00e9d\u00e9ric Pont.<\/p>\n The problem facing Swiss University Hospitals was clear. They needed access to more data, without ever moving, centralizing or revealing patient data.<\/p>\n Data sharing was therefore vital, but at the same time almost impossible. There were three major obstacles:<\/p>\n The hospitals contacted the EPFL (Ecole Polytechnique F\u00e9d\u00e9rale de Lausanne) Data Security Laboratory, led by Professor Jean-Pierre Hubaux, in search of a solution. Aware of the strategic and social importance of the challenge, Professor Hubaux decided to bring in an expert in the form of Juan Ram\u00f3n Troncoso-Pastoriza, specialist in applied cryptography and genomic data protection.<\/p>\n The growing team dedicated themselves to this critical issue, ultimately developing an approach that combines homomorphic encryption and secure multipartite computation so that statistical analyses can be carried out and AI models developed without ever revealing or moving the raw data. \u201cToday, this innovation is called Multiparty Homomorphic Encryption (MHE). It is patented, published, and has been successfully tested on real-world hospital use cases.<\/em>\u201d says Fr\u00e9d\u00e9ric Pont.<\/p>\n Thanks to this pragmatic, strategic leap, institutions can finally work together while maintaining complete sovereignty of their data.<\/p>\nA clear need: sharing data without exposing it<\/h2>\n
\n
Working together without sharing data: EFPL\u2019s contribution<\/h2>\n
The creation of Tune Insight: transforming academic innovation into operational capacity<\/h2>\n