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

The purpose of our present study was to develop a forecasting method that would help asthmatic individuals to take evasive action when the probability of an attack was at THEIR PERSONAL THRESHOLD levels. The results are encouraging. Risk factor analysis helps improve the agent's performance (by allowing it to consider personalized risk score of asthma attack triggers while making a decision and being able to ignore the non-triggers), increasing transparency of deep reinforcement learning in medicine applications (by using the results of analyzing risk factors and its association to take actions), and increase accuracy over time since the association risk factor indicators are also changing over time with more accuracy rate. It also brings the possibility of including population-based health in personalized health, which could support a more efficient self-management of chronic diseases.

Personalized Prediction of Asthma Severity and Asthma Attack for a Personalized Treatment Regimen.

Control of asthma is critical for disease management and quality of life. Asthma treatment depends on the patient demographic information (e.g., age), and disease severity, which is determined by: (1) how symptoms affect a patient's daily life, (2) measured lung function, and (3) estimated risk of having an asthma attack. In this paper, we will present the Tensorflow Text Classification (TC) method to classify a patient's asthma severity level. We will also propose a Qlearning method to train an agent through trials and errors to improve the prediction accuracy and create a personalized treatment regimen for asthma patients.