Privacy-preserving medical image analysis
This work addresses the critical conflict between data usage and privacy protection in medical AI, offering a solution for healthcare providers and researchers to leverage AI while maintaining ethical and legal compliance.
This paper introduces PriMIA, a software framework for privacy-preserving medical image analysis. It demonstrates that a securely aggregated federated learning model achieves significantly better classification performance than human experts on unseen datasets and enables end-to-end encrypted diagnosis without revealing data or models.
The utilisation of artificial intelligence in medicine and healthcare has led to successful clinical applications in several domains. The conflict between data usage and privacy protection requirements in such systems must be resolved for optimal results as well as ethical and legal compliance. This calls for innovative solutions such as privacy-preserving machine learning (PPML). We present PriMIA (Privacy-preserving Medical Image Analysis), a software framework designed for PPML in medical imaging. In a real-life case study we demonstrate significantly better classification performance of a securely aggregated federated learning model compared to human experts on unseen datasets. Furthermore, we show an inference-as-a-service scenario for end-to-end encrypted diagnosis, where neither the data nor the model are revealed. Lastly, we empirically evaluate the framework's security against a gradient-based model inversion attack and demonstrate that no usable information can be recovered from the model.