Adaptive Threshold Determination Based on Entropy in Active Contour without Edge Method for Malaria Parasite Candidate Detection

摘要

Malaria is one of the most serious diseases that often leads to death. The identification of malaria parasites is commonly done by using microscopic images that are divided into two types: thick and thin blood smear. In thin blood smear images, malaria parasite candidates can be easily identified based on blood cell characteristics. However, in the thick blood smear images, the malaria parasite will be difficult to be distinguished from other objects that are also in the blood smear image. There are several approaches to facilitate the identification of malaria parasites in blood smear images, one of which is by utilizing computer-aided diagnosis. This research proposed a scheme to solve the problem of parasite detection in thick blood smear using segmentation approach. The sensitivity of the proposed scheme obtained was 98.04%, indicating that most of parasites in thick blood smear images have been detected; however, the drawback of the sensitivity result is related to the high number of false positive detections. The main objective of this study is to detect the objects in a complete form in the entire slide images using the segmentation approach.