PCD: A privacy-preserving predictive clinical decision scheme with E-health big data based on RNN

Abstract

As large amount of e-health data is generated exponentially, recurrent neural networks (RNN) can be utilized to make predictive clinical decision which is helpful to improve diagnosis accuracy and reduce diagnosis time. However, it is still a challenging task to guarantee the information security and solve the privacy concerns. We design a new Privacy-preserving Predictive Clinical Decision scheme based on RNN, called PCD, that can predict and alert before diseases occur while preserving the privacy of patients. In PCD, we utilize a homomorphic encryption scheme, so no e-health data will be leaked. PCD could resist various security threats. We design a sequential and an averaged RNN model in real-time systems that is capable to improve prediction accuracy. The experimental results illustrate that our scheme achieves high disease prediction accuracy and time efficiency while protecting privacy of patients.