Malaria is categorised as a dangerous disease that can cause fatal in many countries. Therefore, early detection of malaria is essential to get rapid treatment. The malaria detection process is usually carried out with a 100x magnification of thin blood smear using microscope observation. However, the microbiologist required a long time to identify malaria types before applying any proper treatment to the patient. It also has difficulty to differentiate the species in trophozoite stages because of similar characteristics between species. To overcome these problems, a computer-aided diagnosis system is proposed to classify trophozoite stages of
Malaria is an acute (sudden) and chronic disease that can threaten people’s health worldwide. It has been reported that around 229 million cases and 409,000 deaths worldwide were recorded in 2019 due to malaria, which led to an expenditure of US$ 3.0 billion to control and prevent malaria at the global scale, particularly within the African Region, South-East Asia, Eastern Mediterranean, America and Western Pacific [
PF, PK and PV are commonly discovered in Malaysia [
Clinically, visual inspection becomes quite challenging if a microscopic image contains blur effects, unwanted artefacts and a lot of noise in the image [
In the studies of image processing and automated classification procedure using a computer-aided system, many limitations have been reported by previous researchers. Most of the classification process only focused on the infected RBCs and normal RBCs classification systems [
Thus, this study proposes an intelligent classification system for PK, PF, and PV Malaria species for trophozoite stages. The system is developed using image processing techniques, including contrast enhancement, proposed segmentation procedure, feature extraction and classification process using MLP-BR to classify the image.
This paper is arranged as follows: In Section 2, the methodology to develop an automated classification system is introduced. Then, Section 3 describes the application of the method and experimental results are presented and analysed. Finally, Section 4 provides the conclusion of this work.
The system architecture is implemented using five main processes: image acquisition, contrast enhancement, image segmentation, feature extraction, classification, and Graphical User Interface (GUI), as shown in
Malaria samples were prepared by experts from Hospital Universiti Sains Malaysia, Kelantan, Malaysia. Images were acquired using microscope BX41 connected to Olympus XC50 camera under 100X magnification resolution using immersion oil. The capturing procedure of the sample images has been approved by National Medical Research Register (NMRR), with the register serial number of NMRR-16-1434-31673 (IIR). This study only focuses on differentiating between trophozoite stages of PK, PF, and PV. Moreover, 150 thin blood smear images, including 50 images of each PK, PF and PV, have been captured. Images were saved in the bitmap (BMP) format in 800 × 600 color (RGB) ideas.
The contrast enhancement method aims to enhance the image’s contrast and reduce the color consistency problem of thin blood smear images due to the staining process [
Using the HE method, the input color image is corrected according to the intensity value of the reference image. The reference image used in the algorithm is fixed and does not change for any color correction process of other photos. Therefore, the reference image contains the best intensity among all samples.
Image segmentation aims to separate the region of interest from the unwanted regions [ Load the original color image from Subsection (2.2). Convert the image in Step 1 to a grayscale image. Segment the grayscale image from Step 2 into a binary image using Otsu’s method [ Apply erosion operation by using structuring element ‘disk’ to separate the overlapping cells. Segment image in Step 2 using manual thresholding method according to the following equation: Identify which coordinate in Step 5 overlapped with the cell in Step 4. The rest will be deleted. Infected cells together with an object will be displayed. Apply dilation operation in Step 7 by using the similar structuring element ‘disk’ in Step 4 to obtain the original size of an infected cell.
Steps 1 to 3 are the segmentation process of the cell images, whereas Steps 4 to 8 are the process for obtaining infected cells.
During the segmentation process, the overlapping image occurred after applying Otsu’s method. Therefore, the erosion process has been used to separate the overlapping cells. In Step 5, the manual thresholding method set 50 to 110 intensity value has been used to find the coordinate location of malaria parasite in the image. This intensity value has been selected based on observation from a microbiologist for 150 images during the experiment.
Three features have been extracted, namely the size of infected red blood cell (RBC), brown pigment in the parasite and texture feature using the Gray Level Co-occurrence Matrix (GLCM) technique [
The size of infected RBC is based on the sum of white pixel values obtained from the segmented image shown in
The brown pigment in the trophozoite stage of the malaria parasite only exists in PK species. The hue, saturation and value (HSV) color model [
The GLCM method [
All data from the feature extraction process has been analysed using Multilayer Perceptron (MLP) for the classification process. In this study, the Levenberg-Marquardt (LM), Bayesian Regulation (BR) and Conjugate Gradient Backpropagation (CGP) had been used as a learning algorithm to train and test for the accuracy and robustness of the result [
Training algorithm | Training data (90%) | Testing data (10%) |
---|---|---|
97.54 | 96.35 | |
94.37 | 92.94 |
An Intelligent Classification System for Trophozoite Stages in PK, PF and PV Malaria Species has been successfully analysed using image processing and classified using MLP to get the performance of the data. In the image processing method, the process of contrast enhancement, proposed segmentation procedure and feature extraction is used to obtain image characteristics. The experimental results have been performed using MLP-BR to obtain the good accuracy of classification. The proposed system helps classify trophozoite stages of PK, PF and PV with an accuracy of 98.95%.
The authors would like to thank the technologist in Centre for Research Laboratory (CRL), Hospital University Sains Malaysia (HUSM), Kubang Kerian, Kota Bharu, Kelantan, Malaysia its facilities and providing useful information during this research study.