00000ctm a22000004a 4500 UP-99796217608419809 Buklod 20080610145452.0 m |o d | ta 080610s2008 xx d r |||| u| (iLib)UPD-00027935472 DENGII eng DMLUC LG 995 2008 C65 C53 Clarin, Christine T. Location of malaria infection in cells through grammatical evolution of image processing techniques (MICE) by Christine T. Clarin. 2008. xii, 132 leaves Thesis (M.S. Computer Science)--University of the Philippines Diliman. Malaria is an important tropical disease to address. Over 2 billion people are at risk of infection and at least a million people die annually. In addressing malaria, diagnosis is done first and the standard method is blood film examination. However, this is labor and time intensive and there is a need to lessen and speed up the manual analysis of image samples. This thesis developed MICE, a system that uses grammatical evolution (GE) in dynamically generating an image analysis JAVA macro. This macro detects and locates malaria infection in blood smear images without considering prior information on slide preparation or image acquisition. MICE has two modules: Image Processing and Machine Learning. The image processing module follows the general algorithm: (a) identify parasites, (b) segment and count cells and (c) find overlaps (infected cells) among (a) and (b). Various image processing techniques are required to restore the raw images to a standard before performing a, b and c. From these, the system creates a language that preprocesses images then applies morphological operators on it. The grammar of the language is the input to GE. It evolves productions within the grammar to produce a code specific to the image's condition. The results of MICE are competitive with other previous efforts in automation with manual location of infected cells as reference, This indicates that GE with the use of a raw fitness function was effective in giving image specific codes for the detection and location of malaria. The use of ImageJ macros for programming and a simple genetic algorithm as a search engine for genome to phenotype mapping were also sufficient for the success of the system. Overall, MICE has 75% sensitivity, 90% specificity, 78% and 97% positive and negative predictive value for malaria diagnoses. Imaging systems in medicine. Image processing Digital techniques. Malaria. UP FI Thesis UPD DARCHIVES LG 995 2008 C65 C53 UPD DENG-II LG 995 2008 C65 C53 Thesis