Urine sediment overlapped cells segmentation based on hough transform and geometrical feature

The automatic urinalysis analyzer has been widely researched for its great clinical use which aims to count the visible components automatically. Segmentation is the first and most critical part which directly affects the further identification and statistics. Nevertheless, segmenting of overlapped cells still remains a challenging task. This paper presents a novel approach for the urine sediment image segmentation. We first extract the region of interest and detect circular targets based on Hough transform. Meanwhile, centers of these targets are marked. Secondly, we compute the distance of every boundary point to the center of each region for the overlapped cells. At the points which reach minimum distance, we do the segmentation. Our method has proved good efficiency and accuracy especially for overlapped cells. However, further research needs to be done to deal with more complicated adhesion.