Predicting the effect of the environment parameters in asthma patients using Artificial Neural Networks

• Main Authors: Pei-Ying Chen, 陳珮瑩
• Other Authors: Yu-Chuan Li
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/77381213855352661150

碩士 === 國立陽明大學 === 生物醫學資訊研究所 === 99 === Our life is closely related to the environment. Climate change getting fast and unstable, so people are getting climate-related diseases such as asthma. According to World Health Organization (WHO) estimates, 235 million people suffer from asthma globally, and the prevalence of asthma is increasing. Asthma, become a world-wild health problem. We use existing data in hospital and combine with the environment parameters to find an appropriate model for assist asthma patients in asthma attack control. To reach this propose, patients who were diagnosed as asthma (ICD-9 code: 493.*) by physicians from emergency department (ED) and outpatient department (OPD) during January 2007 to December 2009 were selected. According to the patient data, we can classify all the date to high-risk attack date or low-risk attack date, and we use Atmospheric parameters and Air pollution parameters as variables to predict asthma attack. Through the data preprocessing, feature selection and neural network adjust, we can decide the final predict variables and perform the best prediction of the model. Although the performance of the ED model is not good enough (Accuracy: 66.57%, Sensitivity: 0.666, Specificity: 0.601, AUC: 0.682), but the result from the OPD is much better (Accuracy: 75.18%, Sensitivity: 0.752, Specificity: 0.757, AUC: 0.846). In this study, asthma attack could be predicted by the easy-get environment parameters, the accuracy performance is higher than 70%. In future studies, we could apply this model in clinical practice to help asthma patients.