White Blood Cells Detection and Counting Based on Otsu Thresholding and Optimized Elipse Fitting by means of Image Processing

In this research, a new method for detection and counting of white blood cells in blood smear images using image processing methods is proposed. Detecting and counting of white blood cells is very important due to its association with the diagnosis of the health status of individuals and the examination of diseases. There are currently some limitations in detection of white blood cells in researches, which has been improved to a satisfactory level in this research. One of them is segmentation in overlapping images. Also, another current problem is inaccurate detection of the cell boundary, which leads to lower counting accuracy. The method used in this research thesis is Otsu Thresholding and Optimized Ellipse Fitting with high accuracy in the image segmentation and white blood cell detection, especially in overlapping images. In addition to solving the segmentation problem in overlapping images, counting and zoning of each white blood cell are performed more precisely. Considering the geometric shape of the white blood cell, using the ellipse fitting method, the range of each white blood cell is determined and the application of an differential evolution algorithm leads to an optimization of ellipse fitting because multiple ellipses are created around the cell s region until the best ellipse which covers the candidate cell s edge, selected. The validity of the proposed method is high reliability and efficiency by 95.7 % AUC (Area under the ROC Curve), 96.2 % accuracy and 4.62 % miss rate.

Keywords:

White blood cell , ellipse fitting , image processing.