In medical image processing, the segmentation of overlapping nuclei is one of the challenging topics, which relates to its application in diagnostic pathology. To achieve the quantification accuracy (ACC) of the diagnosis, we propose an overlapping nuclei segmentation algorithm using the principle of direction-based flow tracking (DBFT). The DBFT, which consists of direction field preparation and direction field tracking, is performed to provide the direction field and the labeled distinct single nucleus. Its performance is validated with 6375 nuclei from 29 images and compared with two popular overlapping objects segmentation methods, i.e., traditional watershed (TWS) and marker-controlled watershed (MCWS). While the sensitivity (SS) of the DBFT, TWS, and MCWS is 0.981, 0.990, and 0.966, respectively, and the corresponding positive predictive value (PPV) is 0.948, 0.831, and 0.910. The ACC values and F 1 measures obtained from the combination of SS and PPV are used as the total performance measures. While the ACC values from DBFT, TWS, and MCWS are 0.930, 0.824, and 0.882, respectively, the corresponding F 1 measures are 0.964, 0.904, and 0.937. The results clearly show that the DBFT is the best among three methods because it provides the maximum numbers on both ACC and F 1 values.
This paper presents a novel method for cell image segmentation. The approach is based on the application of dynamic force estimation on a static vane (DFESV) to medical image processing problems. The other processes in the approach include smoothing filter using anisotropic diffusion, thresholding using Otsu's algorithm, morphology operation, marking boundary segment, angle image operator, and angle relation merging. Microscopic images from breast cancer tissue are used for evaluation of the approach. Preliminary results show that cell image segmentation based on the DFESV can successfully separate multiple attach cells into distinct single cells.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.