Automatic detection of circulating tumor cells based on microscopic images

Detection of circulating tumor cells (CTCs) plays an important role in early diagnosis of cancer. Traditional detection relies on empirical knowledge of doctors, which is time consuming and suffers from problems such as subjectivity and low repeatability. To improve the objectiveness and efficiency of CTCs detection, an automatic detection method based on digital image processing techniques of scanned microscopic images are proposed in this paper. First, the overall architecture and the image capturing system are introduced. To fully exploit the optical structures of the blood, microscopic images are scanned at ten different focal lengths. Then, an adaptive threshold is proposed for binarization of the images, where morphologic processing operations are applied to detect suspicious CTCs regions. Finally, detection results from all ten layers are fused to generate the final detection output. Location, range and related graphical information are stored in a database to assist further examination, while interactive navigation display is also supported by the system. The effectiveness of the proposed system is verified by simulation experiments.