Identification of Leukemia Cells in Microscopy Digital Image Using Improved Canny Edge Detection Method

Abstract

Leukemia is a blood cancer mutates inside the bone marrow. If unidentified at an early stage, it can lead to severe consequences and later death of a person. To diagnose leukemia, machines and manpower skills are required to identify if the illustrative image is healthy or unhealthy. The manual observation of the large number of images may lead to error in the results. One of the biggest barriers in the result's accuracy is identification of concerned region in the overlapped cells. Motivated by the challenges, framework is developed based on image processing for automatic detection of White Blood Cells in the peripheral blood smear image. In this paper, canny edge detection with Circular Hough Transform is applied to count the number of white blood cells. The outcome is to obtain the classification of the sample images whether the sample is unhealthy or healthy sample image. Total 108 multi-cell Acute Lymphoblastic Leukemia images are considered. It was found that the proposed cell separation method yields an accuracy of 98% in comparison to state of the art segmentation technique.