Classificação de patologias em imagens médicas do sangue.

Abstract

In the present work it is shown the proposal and the development of extraction of images, as well as some improvements proposed in other segmentation techniques and classification, with objective of recognizing pathology in medical images of the blood. The automatic identification of those diseases makes possible the most necessary diagnosis and with larger speed. The approach of the present work follows the stages of segmentation of the images, extraction of the visual characteristics and the classification of the pathology. In the segmentation stage the separation technique is applied by color for obtaining of the red blood cells borders, of the leukocytes and of the platelets. The extraction is the stage where the cells are separate from the image, this extraction is accomplished through proposed procedures and developed for this aim at. Finally, the classification is executed through similarities among the images of the leucócitos, using statistical techniques among spatiograms. The results were shown promising as the viability of the application of the methodology in support systems to the diagnosis for images, due to simplicity of the application of the same ones as well as of the obtained results.