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

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Coronary heart disease (CHD) is a leading cause of mortality globally.
• Current diagnostic methods like coronary angiography are invasive, costly, and carry risks.
• Existing non-invasive data (echocardiography, blood tests, patient demographics) offer partial insights into heart damage.

Purpose of the Study:

• To develop and validate a novel computational model for predicting coronary heart disease (CHD) grade.
• To leverage a combination of non-invasive patient data for improved diagnostic accuracy.
• To offer a less invasive and potentially more cost-effective alternative to coronary angiography.

Main Methods:

• Proposed a combined reinforcement multitask progressive time-series networks (CRMPTN) model.
• Employed deep reinforcement learning (DRL) with asynchronous advantage actor-critic (A3C) for pre-training.
• Utilized recurrent neural networks (RNNs) and parameter sharing modules for multitask learning and prediction.

Main Results:

• The CRMPTN model demonstrated satisfactory performance in predicting CHD grade.
• DRL pre-training enhanced the interaction and collaborative learning among multiple tasks within the model.
• The proposed model outperformed existing state-of-the-art methods in predicting coronary heart disease status.

Conclusions:

• The CRMPTN model offers a promising non-invasive approach for coronary heart disease diagnosis.
• Integrating diverse patient data through advanced AI techniques can improve diagnostic capabilities.
• This method holds potential for wider application, especially in resource-limited settings.