Human Dendritic Cells Segmentation Based on K-Means and Active Contour

Dendritic cells play a fundamental role in the immune system. The analysis of these cells in vitro is a new evaluation technique of the effects of food contaminants on the immune responses. This analysis that remains purely visual is a laborious and time-consuming process. An automatic analysis of dendritic cells is suggested to analyze their morphological features and behavior. It can serve as an assessment tool for dendritic cells image analysis to facilitate the evaluation of toxic impact. The suggested method will help biological experts to avoid subjective analysis and to save time. In this paper, we propose an automated approach for segmentation of dendritic cells that could assist pathologists in their evaluation. First, after a preprocessing step, we use k-means clustering and mathematical morphology to detect the location of cells in microscopic images. Second, a region-based Chan-Vese active contour model is applied to get boundaries of the detected cells. Finally, a post processing stage based on shape information is used to improve the results in case of over-segmentation or sub-segmentation in order to select only regions of interest. A segmentation accuracy of 99.44% on a real dataset demonstrates the effectiveness of the proposed approach and its suitability for automated identification of dendritic cells.