Leukocyte Counting using Combination of First Order Statistical Parameters and Otsu’s Thresholding in Microscopic Blood Image

Leukocyte counting is a routine blood examination that is often requested in health care units and hospitals in an effort to carry out the established diagnosis, to provide information about various conditions of these diseases. A framework for counting leukocytes has been proposed in this study. A thresholding method that combines first order statistical parameters with the otsu function has been applied to extract the nucleus on a microscopic image of blood, followed by marker-controlled watershed method to separate the touching nucleus. Due to the presence of several abnormal leukocyte regions because of the staining process error, the elimination of the abnormal region is performed. In the end, it is completed by the connected component labeling. The experimental results show that the segmentation results have an average sensitivity (true positive rate) of 97.2%, an average specificity (true negative rate) of 99.03%, while the false positive rate reaches 0.97% compared to the groundtruth image. The results of the leukocyte counting experiment showed that the average over counted ratio reached 15.86%, while the under counted ratio of the proposed framework was 1.09%.