A Novel Cascading Red Blood Cell Segmentation Method

作者：Jiang, Peilin; Wang, Fei; Wang, Xing; Zhang, Xuetao; Ding, Ding; Li, Hui; Zhang, Guohe^*

来源：Journal of Medical Imaging and Health Informatics, 2016, 6(7): 1708-1713.

DOI：10.1166/jmihi.2016.1876

• Red Blood Cells
• Segmentation
• Connected Component
• Hough Circle Detection

The analysis of erythrocyte morphology is an important determinant in the diagnosis of many diseases. The development of more reliable and automated image analysis techniques for Red Blood Cells (RBCs) image remains an important goal in computational molecular biology. Due to the characteristic of RBCs image which captured by electron microscope (EM), a novel cascading red blood cells segmentation method is proposed. Firstly the feasibility of threshold segmentation has been analyzed and a method based on the number of connected component for coarse segmentation is put forward, successfully extracts the RBCs areas from the images with accuracy 97.52%. Then an marked watershed algorithm bases on random Hough circle detection and Mathematical Morphology is proposed for overlapped RBCs in original image. Compared to the widely used method, the accuracy of this method has improved obviously. The experimental results show that the proposed method is robust to different RBCs images in automatically separating RBCs, the extraction rate reached 82.17%. At the same time, the automatic counting accuracy up to 96.24%.

• 出版日期2016-11
• 单位西安交通大学