Erythrocytes morphological classification through HMM for sickle cell detection

Abstract

In sickle cell disease the cell morphology analysis is used to diagnose due the deformation of the red blood cell caused by the disease. Previous works used, in images of peripheral blood samples, ellipse adjustment and concave point detection due to the elongated shape of the erythrocyte and obtained good results the detection of cells that were partially occluded in cells’ clusters. In this work, we propose a new algorithm for detecting noteworthy points in the ellipse adjustment and the use of Hidden Markov Model (HMM) for automatic erythrocyte supervised shape classification in peripheral blood samples. Furthermore, in this study we applied a set of constraints to eliminate the image preprocessing step proposed in previous studies. The method was validated using peripheral blood smear samples images with normal and elongated erythrocytes. In all the experiments, in the classification of normal and elongated cells the sensibility was superior to 96%.