Abstract

Info

Title:: Deep learning-based model for risk prediction of cardiovascular diseases

Author(s):: AN Ying1; HUANG Nengjun2; YANG Rong3; CHEN Xianlai1; 1. Information Security and Big Data Research Institute, Central South University, Changsha 410083, China; 2. School of Computer Science and Engineering, Central South University, Changsha 410083, China; 3. Xiangya Hospital, Central South University, Changsha 410078, China

Keywords:: ; Keywords: cardiovascular disease; risk prediction; electronic medical record; deep learning

Abstract:: Abstract: The accurate prediction of cardiovascular diseases (CVD) is of great significance for the prevention of CVD. Therefore, a novel model based on electronic medical records (EHR) and data mining is proposed to investigate the risk prediction of CVD. Recurrent neural network is adopted for the representation learning of EHR, which can effectively capture temporal information hidden in EHR and achieve feature engineering without any manual intervention. Meanwhile, a context vector which is obtained via attention mechanism embed in recurrent neural network model can improve the fitting performance as well as interpretability of the risk prediction model. To further improve the accuracy of the risk prediction of CVD, the model effectively combines various kinds of clinical data, including diagnostic coding sequence, laboratory data and demographic statistics. The model utilizes several modules for representation learning, which can take full consideration of not only the difference but also the correlation among these clinical data, thus improving the performance in the risk prediction of CVD. Experimental results show that the proposed model outperforms the latest methods in the risk prediction of CVD. The recall rate, F1-score and AUC of the proposed model can reach 0.814 9, 0.737 8 and 0.837 5, respectively.