A Novel Approach for Objective Assessment of White Blood Cells Using Computational Vision Algorithms.

In the field of medicine, the analysis of blood is one of the most important exams to determine the physiological state of a patient. In the analysis of the blood sample, an important process is the counting and classification of white blood cells, which is done manually, being an exhaustive, subjective, and error-prone activity due to the physical fatigue that generates the professional because it is a method that consumes long laxes of time. The purpose of the research was to develop a system to identify and classify blood cells, by the implementation of the networks of Gaussian radial base functions (RBFN) for the extraction of its nucleus and subsequently their classification through the morphological characteristics, its color, and the distance between objects. Finally, the results obtained with the validation through the coefficient of determination showed an overall accuracy of 97.9% in the classification of the white blood cells per individual, while the precision in the classification by type of cell evidenced results in 93.4% for lymphocytes, 97.37% for monocytes, 79.5% for neutrophils, 73.07% for eosinophils, and a 100% in basophils with respect to the professional. In this way, the proposed system becomes a reliable technological support that contributes to the improvement of the analysis for identification of blood cells and therefore would benefit the low-level hematology establishments as well as to the processes of research in the area of medicine.