Classification of Candida Cells with Morphological Identification

Abstract

Candida causes fungal infections in human and animals. Since a large number of Candida cells can be found in one sample, manual quantification and classification has become a tedious task for technicians. Moreover, it is used to identify fungal infections and for treatments. Therefore classification and morphological identification of these organisms is an essential feature in the diagnostic process. Live Candida cells appear in purple color and dead cells appear in pink. Since Candida cells appear as overlapped cells, it is really important to separate each touching cells and count them as individuals. Color segmentation can be used to classify live and dead cells. It is achieved by analyzing the red component grayscale image. The well famous watershed method has been used for the segmentation of touching Candida cells. The Distance Transform is used to apply watershed method as it enables any types of intensity images. Introducing a semi-automated system avoids the uncertainties where the user can remove the artifacts from the final result of the system. Also seeded region growing algorithms can be used for further segmentation if the user can select seed points manually.