Increasing Performance of Plasmodium Detection Using Bottom-Hat and Adaptive Thresholding
Chyntia Raras Ajeng Widiawati, Hanung Adi Nugroho, Igi Ardiyanto, M. Syaiful Amin · 2021
Based on WHO data, in 2019 there were around 229 million cases of malaria worldwide, with a death rate of 409,000 people. This disease can be cured if diagnosis is done accurately and quickly. The gold standard for malaria diagnosis is microscopic examination. Despite the huge number of microscopic examinations, there are still some misdiagnose caused by human error. The previous research has developed a computer aided diagnosis (CAD) method for identify the phase of Plasmodium Falciparum. However, the study focused only on thin blood smear, meanwhile in cases of thick blood smear, a better technique is needed because plasmodium in thick blood smear is not well visualized and very small. The purpose of this research is to develop detection method based on digital image processing to support the malaria detection by paramedics. The result of this research will be able to differentiate Plasmodium parasites and artifacts. The data used is a digital microscopic image of thick blood smear. Data acquisition was held in Laboratory of Parasitology, Faculty of Medicine Universitas Gadjah Mada. 846 objects are obtained from 38 image in a time. There are nine features of the texture. The features are then selected by the Wrapper, CFS and the Gain Ratio feature selection method. There are three different classification comparisons in the classification stage, including naïve Bayes (NB), support vector machine (SVM) and multilayer perceptron (MLP). The best results in the classification of species were obtained from the NB classification with an accuracy of 82.05%.