Risk factors identification for heart disease in unstructured dataset using deep learning approach

Abstract

An automatic identification of the heart disease status can give timely support to medical decision-making process. The identified key factors can expedite the prevention actions in right direction. Existing solutions to identify disease factor or current disease status is based on hybrid approach which requires significant amount of human efforts. In addition to that an information extraction and de-identification on clinical dataset performed manually is error prone, expensive, prohibitively and time consuming [1]. These drawbacks can be overcome by using deep learning approach, in this paper we have used LSTM, BiLSTM and Google's Sentence encoder for automatic disease status identification. We have used i2b2 dataset with the proposed deep learning models that have led to promising results.