Cardiovascular diseases (CVD) as a complex disease involves genetic, environmental, and lifestyle causes and the interactions of such factors. How to combine such heterogeneous features into the predictive model remains a challenge. Deep learning models such as convolutional neural networks (CNN) and LSTM can explore temporal patterns on Electronic Health Records (EHR) features to enhance disease prediction. In this study, we utilized a concatenating framework of CNN to combine temporal EHR data and genetic features for prediction. We compared the predictability of this approach by using only EHR. The experimental results showed that the proposed models by adding genetic features offered small but significant benefits to clinical features, which underline the importance of genetic data.
Learning Objective: 1. Learn how to use deep learning on temporal EHR data for CVD prediction
2. Learn how to use a concatenating framework to combine temporal EHR data and genetic features for prediction.
Juan Zhao (Presenter)
Vanderbilt University Medical Center
QiPing Feng, Vanderbilt University Medical Center
Patrick Wu, Vanderbilt University Medical Center
Joshua Denny, Vanderbilt University Medical Center
Wei-Qi Wei, Vanderbilt University Medical Center