R language Web Application Development– Invent “Heat Map” Disease Clustering Analysis for Environment Health Risk Assessment

• Main Authors: Hsin-Ling Yeh, 葉信伶
• Other Authors: Po-Huang Chiang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/49818126243611632603

碩士 === 國立陽明大學 === 生物醫學資訊研究所 === 101 === Changhua and Yunlin are two major agricultural counties in Taiwan, where heavy environmental pollution occurred due to rapid industrialization in the 1970’s. As a result, health of residents in certain areas was severely affected. The study used cluster analysis to comprehend the disease clustering and further explore the correlation between disease and environment. Previous studies of a region cluster state of mortality use Geographic Information Systems (GIS) and spatial analysis. However, these methods only focus on one disease and high cost, while we need to combine all the data to overlay and compare outcomes. Thus we are unable to use this method because the final image from using this type of analysis on our data would complicate the presentation of the results. To facilitate health risk assessment in these areas, this R web tool was developed to identify pollution “hot-spot” areas using a spatial autocorrelation model and heat map with hierarchical clustering in mortality risk. We can then examine the important health risk factors these particular areas. The system was developed with a freeware of R script, Java Server Pages (JSP) that displayed the results of a cluster analysis by permuting the rows and the columns of a matrix of place similar values near each other. This study used GIS spatial autocorrelation analysis and cluster heat map in the R web tool to survey the relationship between the highest death rate local and industrial pollution by analyzing the top ten leading causes of death in Changhua and Yunlin, from 2001-2011.We found the cluster states of male and female individual in Changhua 2006, 2001 and Yunlin 2005, 2008 spatial analysis were significant (Moran’s I, p-value<0.05). Thus we graphically visualized these clusters and found that in these clusters the industrial air pollution had similar distribution to the death rates. In the other part of this tool, temporal analysis used female death rate of cancer in Yunling, 2001-211.