A Deep Learning Approaches for Enhancing Clinical Solutions to Cardiovascular Prediction Using EHR

Abstract

The prediction of cardiovascular disease gained immense significance in the medical field with the alignment of increasing focus on promoting healthier lifestyle. Current methods for cardiovascular disease prediction is leading to so many miss classifications, urging the need of modern automated Deep learning approaches. The main purpose of these approaches is to detect the occurrence of cardiovascular disease (CVD) using patient information from comprehensive electronic health records (HER). Moreover, it is a complex task to choose appropriate features from Electronic Health Records data, and it is a huge confronts to attain robust and accurate results because of the incomplete data entry errors, incomplete record of the patient and patient self-reporting and data integration issues. In this paper we propose an efficient end-to-end framework known as Risk prediction with Deep Residual Neural Network (DRNN), which not only acquires the most influencing features; but also considers the time-based medical data and temporal data to help the patient disease progression, treatment effectiveness, and to check how other diseases are affecting the state of patient. The experimentation is done with the online available Kaggle dataset for cardiovascular disease (CVD) prediction. The result of DRNN demonstrate that the anticipated model significantly enhances the prediction accuracy and F-Measure, Sensitivity compared to various existing approaches. The anticipated model establishes superior trade-off among other approaches.