Unsupervised Leukocyte Image Segmentation Using Rough Fuzzy Clustering

摘要

The segmentation of leukocytes and their components actsas the foundation for all automated image-based hematological diseaserecognition systems. Perfection in image segmentation is a necessarycondition for improving the diagnostic accuracy in automated cytology. Sincethe diagnostic information content of the segmented images is plentiful,suitable segmentation routines need to be developed for better diseaserecognition. Clustering is an essential image segmentation procedure whichsegments an image into desired regions. A judicious integration of roughsets and fuzzy sets is suitably employed towards leukocyte segmentationin a clustering framework. In this study, the goodness of fuzzy setsand rough sets is suitably integrated to achieve improved segmentationperformance. The membership concept of fuzzy sets endow is efficient handlingof overlapping partitions, and the rough sets provide a reasonable solution todeal with uncertainty, vagueness, and incompleteness in data. Such synergisticcombination gives the proposed scheme an edge over standard cluster-basedsegmentation techniques, that is, K-means, K-medoid, fuzzy c-means, and roughc-means. Comparative analysis reveals that the hybrid rough fuzzy c-meansalgorithm is robust in segmenting stained blood microscopic images. Theaccomplished segmented nucleus and cytoplasm of a leukocyte can be usedfor feature extraction which leads to automated leukemia detection.