Accurate Segmentation of Partially Overlapping Cervical Cells Based on Dynamic Sparse Contour Searching and GVF Snake Model
Publication in refereed journal


摘要Overlapping cells segmentation is one of the challenging topics in medical image processing. In this paper, we propose to approximately represent the cell contour as a set of sparse contour points, which can be further partitioned into two parts: the strong contour points and the weak contour points. We consider the cell contour extraction as a contour points locating problem and propose an effective and robust framework for segmentation of partially overlapping cells in cervical smear images. First, the cell nucleus and the background are extracted by a morphological filtering-based K-means clustering algorithm. Second, a gradient decomposition-based edge enhancement method is developed for enhancing the true edges belonging to the center cell. Then, a dynamic sparse contour searching algorithm is proposed to gradually locate the weak contour points in the cell overlapping regions based on the strong contour points. This algorithm involves the least squares estimation and a dynamic searching principle, and is thus effective to cope with the cell overlapping problem. Using the located contour points, the Gradient Vector Flow Snake model is finally employed to extract the accurate cell contour. Experiments have been performed on two cervical smear image datasets containing both single cells and partially overlapping cells. The high accuracy of the cell contour extraction result validates the effectiveness of the proposed method.
著者Guan T, Zhou DX, Liu YH
期刊名稱IEEE Journal of Biomedical and Health Informatics
頁次1494 - 1504
關鍵詞Cervical smear image; dynamic sparse contour searching (DSCS); GradientVector Flow (GVF) Snake model; medical image processing; overlapping cells segmentation
Web of Science 學科類別Computer Science; Computer Science, Information Systems; Computer Science, Interdisciplinary Applications; Mathematical & Computational Biology; Medical Informatics

上次更新時間 2020-25-10 於 02:05