Research On Segmentation Algorithm For Urine Sediment Overlapped Cells

The automatic urinalysis analyzer works to check the components of urine sediment, such as red blood cells(RBCs) and white blood cells(WBCs). However, cell distribution, section making and some other conferring factors may cause overlapping, mostly representing as RBCs’ adhesion and WBCs’ adhesion. If the overlapped cells are not segmented, the final counting and diagnosis is questionable in a sense.How to segment the overlapped cells is the focus of this paper. In the beginning, we discussed the achievements done by other researchers. Then we proposed three novel methods to realize the segmentation.The first method based on Freeman code starts by using the 8-directional chain code to track the cell contour. The concave points are those whose chain code difference is less than the threshold. Then at the points with the maximum ratio of chord and arc, we do the segmentation. It’s proved to be easy to implement, but it only applies to two cells’ adhesion with smooth edges, such as RBCs. For the overlapped WBCs with defects on the boundaries, we use the area method to segment them.In the case of three cells’ adhesion, we can process them serially. But it takes much more computation, so we try the second method to fulfill the segmentation. In short, cells are detected by Hough Transform and the overlapped cells are segmented based on the geometrical feature. This method goes for both RBCs and WBCs, but doesn’t work when there are more than three cells adhering.In order to deal with multiple cells’ adhesion, we proposed the third method, in which we obtain the minimum convex hull by Graham scanning following by using the hull minus cell region to acquire the defects. Then according to the distance of the boundary points of defect to its corresponding cells’ center, the segmentation points are selected. Compared to the approaches mentioned above, this method is more robust and effective for the complex adhesion.