AUTOMATED QUANTIFICATION AND CLASSIFICATION OF MALARIA PARASITES IN THIN BLOOD SMEARS

• Main Author: MOHD AZIZ, SITI SARAH AZREEN
• Format: Final Year Project
• Language: English
• Published: Universiti Teknologi Petronas 2013
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://utpedia.utp.edu.my/10041/1/Automated%20Quantification%20and%20Classification%20of%20Malaria%20Parasites%20in%20Thin%20Blood%20Smears.pdf; http://utpedia.utp.edu.my/10041/

Malaria is one of the life threatening diseases caused by mosquitoes of Anopheles genus that carries the plasmodium parasite. In recent practice, popular methods of malaria disease identification are based on parasitological testing and diagnosis based on symptoms. Both methods have several drawbacks such as limited access to microscopy experts especially in rural area practice, and restricted diagnostic facilities. In addition, accuracy rate is very much dependent in level of microbiologist’s expertise and experience level. Thus, there is an urge for a fast and highly accurate diagnosis technique. The main objective of this project is to improve the current diagnosis technique of malaria parasite in thin blood smears by means of automatic identification by using an image processing method. Focus will be on identifying and counting Plasmodium Vivax parasite at trophozoites stage in thin blood smears. Experiment is conducted in MATLAB environment specifically using the Image Processing Toolbox. Tasks will be divided into four main stages; image acquisition, image preprocessing, image segmentation and image classification. In preprocessing, images were converted to L*a*b* color spaces and are filtered to remove noises. For segmentation stage, a threshold for each image was calculated by using Otsu method. Further, dilation and erosion were performed to completely removed background elements. In the classification stage, images were classified based on the number of infected red blood cell detected. Testing has been done by using 350 images had yield in 99.72% sensitivity, 99.94% specificity and 98.90% positive predictive value. Result proved that this proposed method is able to automatically quantify and classify malaria parasites accurately.