Personalized medicine is set to revolutionize healthcare, yet large-scale research studies towards better diagnoses and targeted therapies are currently hampered by data privacy and security concerns. New global collaborative research has developed a solution to these challenges, described in Nature Communications.
Predictive, preventive, personalized and participatory medicine, known as P4, is the healthcare of the future. To both accelerate its adoption and maximize its potential, clinical data on large numbers of individuals must be efficiently shared between all stakeholders. However, data is hard to gather. It’s siloed in individual hospitals, medical practices, and clinics around the world. Privacy risks stemming from disclosing medical data are also a serious concern, and without effective privacy preserving technologies, have become a barrier to advancing P4 medicine.
Existing approaches either provide only limited protection of patients’ privacy by requiring the institutions to share intermediate results, which can in turn leak sensitive patient-level information, or they sacrifice the accuracy of results by adding noise to the data to mitigate potential leakage.
Now, researchers from EPFL’s Laboratory for Data Security, working with colleagues at Lausanne University Hospital (CHUV), MIT CSAIL, and the Broad Institute of MIT and Harvard, have developed “FAMHE”. This federated analytics system enables different healthcare providers to collaboratively perform statistical analyses and develop machine learning models, all without exchanging the underlying datasets. FAHME hits the sweet spot between data protection, accuracy of research results, and practical computational time - three critical dimensions in the biomedical research field.
In a paper published in Nature Communications on October 11, the research team says the crucial difference between FAMHE and other approaches trying to overcome the privacy and accuracy challenges is that FAMHE works at scale and it has been mathematically proven to be secure, which is a must due to the sensitivity of the data.
Find out more here.