Authors : Dhanya Bibin and P. Punitha
Abstract: This study proposes an efficient approach for automatic detection of clumped cells in peripheral blood smear images that combines Chan Vese segmentation and ellipse fitting. Clumped cell detection is a critical task in developing efficient blood cell segmentation methods in automated malaria diagnosis systems. One of the major challenges in the segmentation of blood cell images is the presence of closely clumped cells. The proposed clumped cell detection method has three steps. Initially, the image segmentation is performed using Chan Vese algorithm to extract clumped and non clumped blood cells from the background. Then, two post-processing techniques are applied on the segmented image to eliminate artefacts, platelets and holes which are irrelevant in this study. The artefacts and platelets are removed using a size based thresholding method and hole elimination is achieved using morphological reconstruction. Then, a robust and efficient ellipse fitting is performed on the connected components in the segmented image to differentiate them as clumped cells or non-clumped cells. The differentiation is based on a threshold value which is computed from the eccentricity of the fitted ellipse. The robustness and efficiency of the proposed method has been evaluated by comparing empirically to manual detection performed by haematologists.
Dhanya Bibin and P. Punitha, 2017. Segmentation and Clumped Cell Detection in Microscopic Peripheral Blood Smear Images. Journal of Engineering and Applied Sciences, 12: 9496-9501.