There are innumerable medical insights lying within the health records of all the people on Earth. This massive amount of data is something for which machine learning applications are perfectly suited: By analyzing data and looking for patterns, AI has been able to learn how to detect diseases and identify genes that seem to play a role in causing them.
But the main problem is access. How can machine learning AI be put to use if the majority of health data is locked behind HIPAA and other regulations that help protect patients’ private health information?
At the Frontier Development Lab, an AI accelerator, researchers are upending the conventional structure of machine learning applications in order to derive medical insights without breaching patient privacy. Beyond solving the privacy problem, the approach could also bolster the reliability of medical datasets, helping researchers find cures to rare diseases.