Abstract

HEROLD-GARCIA, Silena et al. Red Blood cell classification with hidden Markov models. RCIM [online]. 2016, vol.8, suppl.1, pp.499-514. ISSN 1684-1859.

A study of the performance of Hidden Markov Models (HMM) in morphologic supervised classification of erythrocytes in peripheral blood smears of patients with sickle cell disease is realized. Contours are represented in original way considering the angular differences in the curvature of the points of the same. The training of every model comes true with the normal description of the contours and with the representation of the rotation of the same, in order to guarantee a bigger stability in the esteemed parameters. A process of validation crossed of 5x1 for estimate of the error is developed. The measures of sensibility, precision and specificity of classification are obtained. The best results obtain when classifying erythrocytes in two classes, with sensibility values in normal of 96 % and elongated 99 %. In the classification of erythrocytes considering the class of other deformations better results obtain accomplishing the training of the models with the rotation of all the contours, that it attained sensibilities of normal (94 %), elongated (82 %) and with other deformations (76 %).

Keywords : erythrocytes morphological classification; hidden Markov models; contour representation.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )