Document Type : Research Paper

Authors

• Suhad Al-Shoukry ¹
• Zalili Binti Musa ²
• Duha Amer ¹
• Safaa Muhsen Kareem ¹

¹ Department of Computer Systems Techniques, Al-Najaf Technical Institute, Al-Furat Al-Awsat Technical University (ATU), Kufa, Al-Najaf, Iraq

² University Malaysia Pahang (UMP), Pekan, Pahang, Malaysia

Abstract

Image segmentation is an important part of image processing, and one of the most common approaches is threshold segmentation. A new segmentation technique with each pixel in the image has its own threshold is developed in response to the fact that standard threshold-based segmentation algorithms only establish one or many thresholds, making it difficult to extract the complex information in an image. This work employs image segmentation tools to examine images of thin blood smears data set. The goal is to explore options for a noniterative-based and automated system for detecting parasites in blood smears. This can be achieved by detecting the presence of a parasite in thin blood smears and quantifying the portion of red blood cells in the sample that are infected. First, we try segmenting the individual red blood cells from the background using the color thresholder. Next, we clean up the obtained cell mask and examine cell properties using the image region analyzer function, which allows quickly filling in region holes and filtering out regions based on their properties such as area dimensions or eccentricity. Then quickly gauge and specify the expected diameter range of the cells in pixels and indicate that the circles are dark relative to the background. Finally, we've combined the code for finding circles matching image histograms and the parasite threshold detection logic into a single function to quickly examine the performance of this function on the other images using the image batch processing technique. The proposed detection function labels the detected cells with blue circles the parasites are marked in red and the infected cells are highlighted in green. The proposed algorithm has appropriately compensated for the variability in image quality.

Keywords

• Blood Sample Analysis
• image analysis
• blood cell segmentation
• cell morphology
• classification