KLASIFIKASI CITRA MIKROSKOPIK PARASIT MALARIA BERDASARKAN FITUR LUAS DAN TEKSTUR MENGGUNAKAN K-NEAREST NEIGHBOR (KNN)
MetadataShow full item record
Various studies have been carried out to build a malaria parasite type detection system to facilitate the treatment of malaria sufferers by using GLCM feature extraction or using combination of 3-4 feature extractions as a classification input system. Therefore, in this study the aim is to build a system using feature extraction has never been used before, which is a combination of extraction texture features and parasite area features resulting from segmentation. The classification system built in this study uses 3 methods: segmentation method using K-Means to get a value of 1 area feature, feature extraction method using the Gray Level Co-Occurance Matrix (GLCM) to get a value of 16 texture features, and K-Nearest Neighbor (KNN) type of fine KNN used as a classification algorithm. In this study using images of malaria parasites with a total of 90 images gametocyte phase consisting of 30 images of plasmodium falciparum parasites, 30 images of plasmodium malariae parasites, and 30 images of plasmodium vivax parasites obtained from Universiti Sains Malaysia Hospital. The analysis conducted in this study consisted of 2 types, that is the qualitative analysis used in image segmentation using the K-Means method, by looking at the success of the system to separate parasitic cells from normal and background cells. Meanwhile, quantitative analysis is used in the classification system, which is by looking at the calculation results of the classification system accuracy. The results of the analysis carried out in the segmentation process are as many as 86 parasitic images successfully separated from normal cells and backgrounds, and 4 other images are not segmented perfectly due to the presence of normal cells in the segmentation results. While the results obtained in the classification process of 90 images of malaria parasites succeeded in classifying the parasitic images of P. Falciparum as many as 24 images, P. Malariae as many as 30 images, and P. Vivax as many as 25 images with an average accuracy of 87.78%.