Detection of Malaria Parasites using Thresholding in RGB, YCbCr and Lab Color Spaces

In Indonesia, malaria remains an endemic disease and a public health burden. Thus, the Indonesian Government has set a national priority program to eliminate malaria. The goal is to increase the percentage of laboratory-confirmed malaria suspects. In this study, a computer-aided diagnosis is presented to detect and count the Plasmodium parasites in a thick blood smear. The malaria parasite can be detected using the color features. In the thick blood smear, the malaria parasite has purplish or dark red color. Therefore, this study utilizes a color-based segmentation to detect the parasite in three color spaces, i.e. RGB, YCbCr, and Lab. Furthermore, image-based object counting is introduced to count the number of parasites. The best results are achieved by color-based segmentation in RGB color space. In total, there are 27,558 images consisting of infected (13,779) and uninfected (13,779) blood smears. The threshold values are set to 105–240, 10–80, and 75–175 for R, G, and B channels. Of all experiments, the accuracy, sensitivity, and specificity achieved using this scheme are 94.75%, 96.02%, and 93.47%.