New research derives an AI-based method to protect the privacy of medical images.

On May 24th, researchers from the Technical University of Munich (TUM), Imperial College London, and OpenMined, a non-profit organization published a paper titled "End-to-end privacy-preserving deep learning on multi-institutional medical imaging."

The research unveiled PriMIA- Privacy-Preserving Medical Image Analysis that employs securely aggregated federated learning and an encrypted approach towards the data obtained from medical imaging. As the paper states, this technology is a free, open-source software framework. They conducted the experiment on pediatric chest X-Rays and used an advanced level deep convolutional neural network to classify them.

Although there exist conventional methods to safeguard medical data, they often fail or are easily breakable. For example, centralized data sharing methods have proved inadequate to protect sensitive data from attacks. This nascent technology protects data by using federated learning, wherein only the deep learning algorithm is passed on while sharing the medical data and not the actual content. They also applied secured aggregation, which prevents from external entities finding the source where the algorithm was trained. This will not allow anybody to identify the institution where it originated, keeping the privacy intact. The researchers also used another technique to ensure that statistical correlations are derived from the data records and not the individuals contributing the data.

According to the paper, this framework is compatible with a wide variety of medical imaging data formats, easily user-configurable, and introduces functional improvements to FL training. It increases flexibility, usability, security, and performance. "PriMIA's SMPC protocol guarantees the cryptographic security of both the model and the data in the inference phase," states the report.

A report by the Imperial College London quotes professor Daniel Rueckert, who co-authored the paper and says, "Our methods have been applied in other studies, but we are yet to see large-scale studies using real clinical data. Through the targeted development of technologies and the cooperation between specialists in informatics and radiology, we have successfully trained models that deliver precise results while meeting high standards of data protection and privacy."

With the advent of technology and the rapid adoption of AI, the healthcare sector has been witnessing a digital boom. With electronic health records and the proliferation of telemedicine, there is an abundance of medical data and images generated each day. To enable better patient monitoring, diagnostics, and availability of data, these medical data are often shared across different points and institutions. This AI-driven privacy-preserving technology has a potential role to play here as it does not compromise data privacy while sharing happens. And, data cannot be traced back to individuals, thus protecting their privacy.