Abstract
The development of Malaria Parasite Detection and Classification Application can be used in drug susceptibility test for Malaria anti-infection drug development process. The application takes Giemsa-stained thin blood films images as input. The pre-processing is used to enhance image quality. Then, we use local adaptive thresholding to segment the erythrocytes and morphological process to eliminate noises and debris. Watershed transform is used to separate touching and overlapping cells. The labeling matrix is used for localizing each erythrocyte. Texture analysis is performed to establish parameters such as statistical analysis, wavelet transform, gray-level co-occurrence matrix, and gray-level run length matrix. We use minimum redundancy and maximum relevance method with feed forward selection technique to find appropriate parameters for classification of normal cells and infected cells, both ring and trophozoite stages. Then SVMs were used to classify the cells. The algorithms provide 98.77% accuracy for segmentation, 98.41% accuracy for ring stage classification, and 98.99% accuracy for trophozoite stage classification. The application exhibits positions of infected cells and its stage of infection via GUI with %Parasitemia, which can be recorded as excel file format.
โปรแกรมสำหรับตรวจจับและแยกชนิดเชื้อมาลาเรีย
