การจำแนกระยะของเซลล์มาลาเรียชนิดพลาสโมเดียม ไวแว็กซ์ ในภาพฟิล์มเลือดแบบบางโดยใช้ระบบฟัซซีอินเฟอร์เรนซ์

Abstract

Malaria can be spread to human by anopheles female mosquitoes. There are four prevalent Plasmodium species responsible for malaria in humans. One of the four species, Plasmodium vivax or P. vivax, is selected in this paper. Its infected cell detection system based on image processing is developed. We automatically generate membership functions of inputs using four clustering methods were employed: Fuzzy C-Means (FCM), Possibilistic C-Means (PCM), Possibilistic Fuzzy C-Means (PFCM), and Unsupervised Possibilistic Fuzzy C-Means (UPFCM). The Wang-Mendel (WM) method is used to automatically generate rules in the Mamdani fuzzy inference system. In this thesis, three models are developed to classify infected cells at different stages. The first model classifies the Ring stage from Trophozoite, Schizont, and Gametocyte. The second model classifies Trophozoite from Schizont and Gametocyte. Finally, the third model classifies Schizont from the Gametocyte stage. The accuracy of the training data classification for Model 1, which classifies Ring stage cells, achieved the best result of 83.96% using PFCM clustering. Model 2, which classifies Trophozoite stage cells, achieved the best result of 79.71% using PFCM clustering. Model 3, which classifies Schizont stage cells, achieved the best result of 48.61% using PCM, PFCM and UPFCM clustering, while the best result for classifying Gametocyte stage cells was 85.96% using FCM clustering. For the test data, the best classification accuracy for Ring stage cells was 90.48% using PFCM clustering. The best classification accuracy for Trophozoite stage cells was 57.14% using FCM, PCM, PFCM and UPFCM clustering. The best classification accuracy for Schizont stage cells was 87.5% using PFCM clustering, and the best classification accuracy for Gametocyte stage cells was 100% using FCM, PCM, PFCM, and UPFCM clustering.