White Blood Cell Analysis using Different Segmentation Methods on Blood Image Simplified By SMMT Operator
White Blood Cell (WBC) counting plays a major role in the determination of the patientÃ¢â‚¬â„¢s health for different stages, such as diagnosis and treatment. The traditional differential counting method for WBC count is tedious and time consuming. In Indian scenario, major of pathologists use manual methods, for counting the blood cell using microscopic blood image. To automate this process, many have used image processing algorithms such as image segmentation, thresholding, histogram equalization etc. where each technique having some advantages along with some drawbacks at each stage. In this direction, this paper presents WBC segmentation using different image segmentation methods such as watershed transform and level set method for nucleus segmentation and Mathematical Morphology(MM) operator and Granulometric analysis for cytoplasm segmentation Self-dual Multiscale Morphological Toggle (SMMT) operator is used as preprocessing algorithm for image simplification. Total of 46 non-overlapping and 10 overlapping WBC images obtained from internet and tested using four segmentation methods. For nucleus segmentation watershed transform has resulted into 87% of accuracy for non-overlapping images & 50% for overlapping images. The accuracy of watershed transform for non-overlapping images is 10/20% more than that of granulometric analysis method results. Also for cytoplasm segmentation MM Operator is 58% more accurate than Granulometric analysis for non-overlapping images. However Granulometric analysis is found to be 10% more accurate than MM Operator for overlapping WBC images. Watershed transform gives 10.9% more accuracy than level set method for nucleus segmentation whereas MM operator is 58% more accurate than Granulometric analysis for cytoplasm segmentation.