Segmentation of Plasmodium using Saturation Channel of HSV Color Space

Malaria is a vector-borne disease that is spread throughout all regions, especially in the tropical countries. Early malaria detection is necessary to reduce the mortality rate. Microscopic imaging technique is still considered as gold standard for diagnosing malaria. However, due to numerous number of patients with limited parasitological experts, it is necessary to have such an automated system for diagnosing malaria. To obtain more accurate diagnosis, it is essential to have an accurate segmentation of Plasmodium. This study aims to develop a scheme to segment Plasmodium in digital thin blood smear images with different characteristics. There are three main steps consisting of preprocessing, grouping of image characteristics and segmenting Plasmodium. Saturation channel is extracted from HSV in the preprocessing. Difference of image standard deviation is keys for grouping the image characteristics. There are three groups of image that is segmented in different ways. Otsu thresholding combined with morphological image processing is used to segment Plasmodium. The data used contains 124 cropped images. The evaluation results achieve 97.99%, 82.23% and 99.33% of accuracy, sensitivity and specificity, respectively. These high evaluation values indicate that the proposed method is suitable to support the development of CAD in malaria diagnosing.