Nuclei segmentation in microscopy cell images using bayesian clustering and ellipse fitting

Abstract

In recent years, microscopy cell image analysis systems have played a significant part in a wide range of areas such as quantitative pathology, cell biology, pharmacology, and so on. In these systems, cell nuclei segmentation is one of the most essential and important tasks. Despite enormous research efforts and advances, cell nuclei segmentation still remains a big challenge for current cell image analysis systems due to the complex nature of cell nuclei. In a fully automatic cell nuclei segmentation process, one of the main issues to overcome is the problem related to segmenting ``cell clusters" (i.e., overlapped nuclei) since such nuclei will often affect the quantitative analysis of cell images. In this thesis, we present an unsupervised Bayesian clustering scheme for separating cell clusters. The proposed approach first involves applying the distance transform to overlapped nuclei. The topographic surface generated by distance transform is viewed as a mixture of Gaussians in the proposed algorithm. In order to learn the distribution of the topographic surface, the parametric expectation-maximization (EM) algorithm is employed. Cluster validation is performed to determine how many nuclei are overlapped. Our proposed approach incorporates a priori knowledge about the regular shape of single cell nuclei to yield more accurate nuclei delineation results. Extensive experiments have been conducted on a large scale challenging dataset of real-world cell images. Our results demonstrate that the proposed method yields superior segmentation performance, compared to existing state-of-the-art approaches.