Semi-automatic red blood cells counting in microscopic digital images

Abstract

The main purpose of this study is to employ the modern technologies and techniques to semi-automate the quantification process of the Red Blood Cells in Microscopic thin blood smear digital images. The process needs to be more accurate, efficient and universal then the currently practiced methods. The study considers the process to be semi-automated for two reasons, i.e. due to the critical aspect life and due to the diverse nature of the Red Blood Cells in cluster formation. The Methodology of this study involved interactive simple cuts and morphological operations for splitting clusters of Red Blood Cells while counting is carried out through labeling matrix. The Red Blood Cells counting is part of the complete blood count test and is frequently suggested by the Physician to know the number of Red Blood Cells in the patient’s body. The proposed method considers for counting process of the Red Blood Cells first split the clusters and then count the Red Blood Cells. The proposed method achieved an overall True Positive Rate (TPR) of 0.997%, True Negative Rate (TNR) of 0.00265%, accuracy of 0.998% and average error rate of 0.001375% tested on 50 images, data set also on the same number of images linear correlation coefficient R2 is 0.997 between manual and semi-automatic counting of Red Blood Cells.