A Study on Automatic Segmentation of Red Blood Cells in Digital Holographic Microscopy

Digital holographic microscopy (DHM) can provide three-dimensional information. For this reason, this technology has been applied to various fields such as malaria diagnosis using red blood cell (RBC)s and automatic cancer diagnosis using living cells. In research on automatic diagnosis using deep learning, the quality of training data is very important to improve performance. Therefore, we propose an accurate cell segmentation technology for automatic diagnostic assistance of DHM in this paper. The proposed method uses a segmentation algorithm, Segment Anything, and an image recognition algorithm, Xception. In this paper, to evaluate the proposed method, we use five images containing multiple RBCs and calculate the detection rate and false detection rate. This method can reduce the cost of data acquisition and obtain accurate data by automatically acquiring RBCs from images containing multiple RBCs, making it suitable for the automatic diagnosis of DHM using deep learning.