Date of Award


Document Type



Edith Cowan University

Degree Name

Doctor of Philosophy


School of Engineering


Faculty of Computing, Health and Science

First Supervisor

Associate Professor Ute Muelier

Second Supervisor

Associate Professor James Cross


Cancer is one of potentially preventable and treatable diseases. Cancer analysis from different perspectives is necessary to provide the information for health research and the initiation of prevention and treatment programs. The purpose of this study was to analyse five top cancers in the Perth metropolitan area, including lung, melanoma, breast, prostate and colorectal cancers, using two methodologies: Area-to-Point Poisson (ATP) kriging and fitting an inhomogeneous Poisson process model using the Berman-Turner algorithm. ATP Poisson kriging was used to undertake the analysis on the spatial distribution of cancer rates per 100,000 person-years for Perth Statistical Local Areas during the period 1990-2005. This analysis is based on age-adjusted and age-sex-adjusted rates (non-sex specific cancers only). For lung, colorectal and melanoma cancer incidence, it is also evaluated for the period 1990-2005 by sex to investigate how these cancers affect males and females differently. For mortality data, three contiguous periods: 1990-1995, 1996- 2000 and 2001-2005 are studied to investigate how cancers vary over the time. It is demonstrated that female cancer incidence varies more continuously than males and male incidence rates are relatively high. There is difference in estimates and spatial variation between age-adjusted rates and age-sex-adjusted rates. The cancer risk calculated from age sex- adjusted rates shows less variability than the risk only from age-adjusted rates. Lung cancer incidence rates are high in Kwinana industrial area and melanoma cancer incidence rates are relatively high along west coast of Perth. For breast cancer and prostate cancer, the analysis during three contiguous periods is explored. Breast cancer incidence increased gradually with time while prostate cancer mortality decreased with time. Prostate cancer mortality risk from the age is possibly more continuous than the incidence risk from the age. It is believed that cancer mortality rate varies more continuously than the corresponding cancer incidence and that spatial correlation is relatively strong for mortality compared with the incidence. Spatial point pattern theory focuses on the analysis of the incidence (mortality) rates in the year 2005. For each individual cancer adequately fitting models were determined and validated. The model fitting is based on the inhomogeneous Poisson nature of the cancer data. The spatial trends based on the population distribution in Perth metropolitan area are important in obtaining an adequate fit. Except for the spatial trends, the fitted intensity function for the point process is also related to the spatially susceptible population. It is demonstrated that the distance from Kwinana industrial area (KIA) is an appropriate covariate for lung cancer incidence and the distance from the coast accounts for melanoma incidence. Lung cancer incidence increases with proximity to the KIA and similarly melanoma incidence increases with proximity to the coast. The incidence point pattern in 2005 is quite similar to the point pattern in other individual year (or years) when population density is included. Compared with incidence, one covariate (susceptible population intensity) is enough to achieve an adequate fit in most cases for mortality data fitting. The percentage of selected age groups is a reasonable covariate for prostate cancer and lung cancer while population density is a good covariate for other cancers since age has more influence on these two cancer incidences than other cancers. The work shown here presents spatial maps of incidence and mortality risk for the five major causes in Perth, potentially aiding planning.