Deep Q-learning for Predicting Asthma Attack with Considering Personalized Environmental Triggers' Risk Scores

Area of Science:

• Computational medicine
• Artificial intelligence in healthcare
• Personalized medicine

Background:

• Asthma management requires timely intervention to prevent attacks.
• Existing methods lack personalization for individual trigger thresholds.
• Predictive modeling can enhance proactive asthma self-management.

Purpose of the Study:

• To develop a novel forecasting method for asthma attacks.
• To enable asthmatic individuals to take evasive action at personal risk thresholds.
• To improve the accuracy and transparency of predictive models in asthma care.

Main Methods:

• Utilized risk factor analysis to identify personalized asthma attack triggers.
• Employed deep reinforcement learning for decision-making in forecasting.
• Incorporated dynamic updates of risk factor associations over time.

Main Results:

• The developed forecasting method shows encouraging performance.
• Risk factor analysis improved agent decision-making by considering personalized risk scores.
• Increased transparency in deep reinforcement learning applications for medicine was achieved.

Conclusions:

• Personalized risk factor analysis enhances asthma attack prediction accuracy.
• The method supports efficient self-management of chronic diseases like asthma.
• Integration of population health data into personalized health strategies is feasible.