應用空間聚類方法於環境風險區域之篩選

Applying Spatial Cluster Method in Environmental Risk Screening

徐瑋聰(Wei-Tsung Shu)

指導教授 ： 林宏嶽

朝陽科技大學/理工學院/環境工程與管理系/碩士(2008年)

摘要

近年來隨著科技的進步，雖然帶動了台灣經濟的提昇，但同時也造成了嚴重的環境破壞和汙染，並且也直接對人體的健康造成危害。然而，造成的危害並非一時所能查覺，往往都在長時間的累積下才被發現，也已經造成大眾健康與環境無法彌補的損失。因此如何提早發現，成為重要之課題，本研究因此藉由病例資料之空間分佈加以探討，藉由病例資料之異常分佈，找出可能有較高風險之區域。本研究之方法主要係利用不同空間聚類分析步驟進行篩選，以找出案例區域中具有顯著異常疾病發生率之地區，並進一步比較不同方法之差異性、優缺點和適用性，所使用之空間聚類方法可分為網格式分析方法、覆蓋局部案例比例(Overlapping Local Case Proportions, OLCP)與空間掃描統計(Spatial Scan Statistics, SSS)等方法。使用上述之方法，可判斷某罹病率顯著且異常之區域。此外，亦可將上述之數值作為後續自我組織映射圖(Self Organizing Map, SOM)網路之分類因子，藉由SOM物以類聚的特性，分析區域之風險因子與疾病發生率之關係，找尋可能之風險因子。本研究以中部區域做為案例示範區域，以了解上述方法之可行性與應用性。

關鍵字

自我組織映射圖網路 ； 覆蓋局部案例比例 ； 空間掃瞄統計

並列摘要

In recent years, advances in technology not only enhance Taiwan''s economy, but also causing serious environmental impacts and pollution which directly damage public health. However, it is hard to detect the harm caused by the pollution in a short time, consequently causes loss unable to compensate both in environments and public health. Thereby, it becomes important to find the hot spot area early. In this study, the disease data and its spatial distribution are explored to find the probable high risk area where the disease spatial distribution is statistically significant different with other area. The methodology is to apply three spatial cluster analytical procedures to screen the entire case area, including grid analysis method, overlapping local case proportions (OLCP) and spatial scan statistics (SSS). The area whose disease prevalence is significantly different with others will be located after these methods. The advantages / disadvantages, characteristics of these methods are also analyzed. Furthermore, the results of aforementioned spatial cluster methods can be used as input data of self organizing map (SOM) which is a model of artificial neural network, usually being employed for cluster analysis. In this study, SOM can be implemented to analyze the likelihood between candidate risk factors and spatial disease prevalence and find the probable risk factors. In this study, a case study in middle Taiwan is also demonstrated to realize the feasibility and satisfaction of the methodology.

並列關鍵字

self organizing map ； spatial scan statistics ； overlapping local case proportions

參考文獻

1. Best, N. G., Ickstadt, K., Wolpert, R. L., and Briggs, D. J., “Chap 22 Combining models of health and exposure data: the SAVIAH study,” Spatial Epidemiology: Methods and Applications, Edited by Elliott, P., Wakefield, J., Best, N. and Briggs, D., Oxford University Press, USA(2001).

2. Brody, H., Rip, M. R., Vinten-Johnson, P., Paneth, N., and Rachman, S., “Map-making and myth-making in Broad Street: The London cholera epidemic, 1854.” The Lancet, Vol. 356, No. 9223, pp. 64-68(2000).

3. Colten, C. E., and Skinner, P. N., “The road to Love Canal: Managing industrial waste before EPA,” Austin: University of Texas Press, USA(1996).

7. Karakitsios, S. P., Delis, V. K., Kassomenos, P. A., and Pilidis, G. A., “Contribution to ambient benzene concentrations in the vicinity of petrol stations: Estimation of the associated health risk,” Atmospheric Enviornment, Vol. 41, No. 9, pp. 1889-1902(2006).

8. Kulldorff, M., and Nagarwalla, N., “Spatial disease clusters: detection and inference,” Statistics in Medicine, Vol. 14, pp.799-810(1995).