Methodology of high accuracy, sensitivity and specificity in the counts of erythrocytes and leukocytes in blood smear images

Abstract

The evaluation of human blood is an important diagnostic method for the detection of diseases. The analysis of the erythrocytes contained in the blood contributes to the detection of anemia and leukemia, whereas the leukocyte analysis allows the diagnosis of inflammation and/or infections. The blood is analyzed through of the complete blood count test (CBC), which is dependent on automated and/or manual methodologies. The dependence of medical areas on new technologies leads the present study to the goal of developing an image segmentation algorithm that meets the criteria of efficiency and reliability for detection and counting of blood cells. The algorithm was developed through the Matlab software, being the image processing methodology based on the union of the Watershed transform and Morphological Operations, originating the WT-MO methodology. For the simulations, 30 blood smear images containing erythrocytes and leukocytes were used in a non-pathological state. The results showed that the WT-MO methodology presents high sensitivity (99%), specificity (96%) and accuracy (98,3%) when compared with the manual methodology. Therefore, the WT-MO methodology is an accurate, reliable and low-cost technique and can be applied as a third more accessible methodology to perform of the complete blood count test (CBC) in populations of underdeveloped and developing countries.