Prediction of coronary heart disease based on combined reinforcement multitask progressive time-series networks

Abstract

Coronary heart disease is the first killer of human health. At present, the most widely used approach of coronary heart disease diagnosis is coronary angiography, a surgery that could potentially cause some physical damage to the patients, together with some complications and adverse reactions. Furthermore, coronary angiography is expensive thus cannot be widely used in under development country. On the other hand, the heart color Doppler echocardiography report, blood biochemical indicators and personal information, such as gender, age and diabetes, can reflect the degree of heart damage in patients to some extent. This paper proposes a combined reinforcement multitask progressive time-series networks (CRMPTN) model to predict the grade of coronary heart disease through heart color Doppler echocardiography report, blood biochemical indicators and ten basic body information items about the patients. In this model, the first step is to perform deep reinforcement learning (DRL) pre-training through asynchronous advantage actor-critic (A3C). Training data is adopted to optimize the recurrent neural network (RNN) that parameterizes the stochastic policy. In the second step, soft parameter sharing module, hard parameter sharing module and progressive time-series networks are used to predict the status of coronary heart disease. The experimental results show that after DRL pre-training, the multiple tasks in the model interact with each other and learn together to achieve satisfactory results and outperform other state-of-the-art methods.

Keywords: Asynchronous advantage actor-critic; Combined reinforcement multitask progressive time-series networks; Coronary heart disease prediction; Hard parameter sharing; Soft parameter sharing.