Investigation of different classification models to determine the presence of leukemia in peripheral blood image

Abstract

The counting and classification of blood cells allows the evaluation and diagnosis of a vast number of diseases, such as the ALL - Acute Lymphocytic Leukemia, detected through the analysis of white blood cells (WBCs). Nowadays the morphological analysis of blood cells is performed manually by skilled operators, involving numerous drawbacks, such as slowness of the analysis and a non-standard accuracy, dependent on the operator skills. In literature there are only few examples of automated systems able to process a whole image in order to analyze and classify all the WBCs included. This paper presents a complete and fully automatic method for WBCs identification from microscopic images and an evaluation of different classification model to determine the presence of leukemia. Experimental results show that the proposed method is able to identify the cells carrying leukemia and consequently to determine whether a patient is suffering from this disease. © 2013 Springer-Verlag.