Constrained watershed method to infer morphology of mammalian cells in microscopic images

Abstract

Precise information about the size, shape, temporal dynamics, and spatial distribution of cells is beneficial for the understanding of cell behavior and may play a key role in drug development, regenerative medicine, and disease research. The traditional method of manual observation and measurement of cells from microscopic images is tedious, expensive, and time consuming. Thus, automated methods are in high demand, especially given the increasing quantity of cell data being collected. In this article, an automated method to measure cell morphology from microscopic images is proposed to outline the boundaries of individual hematopoietic stem cells (HSCs). The proposed method outlines the cell regions using a constrained watershed method which is derived as an inverse problem. The experimental results generated by applying the proposed method to different HSC image sequences showed robust performance to detect and segment individual and dividing cells. The performance of the proposed method for individual cell segmentation for single frame high-resolution images was more than 97%, and decreased slightly to 90% for low-resolution multiframe stitched images. ' 2010 International Society for Advancement of Cytometry Key terms microscopic image sequence; hematopoietic stem cell; cell uropodia; automated cell shape analysis; watershed segmentation; inverse problem; biomedical image analysis ADVANCED techniques in digital image processing and pattern recognition can potentially be applied to a large number of digital cytometry systems to improve our understanding of cellular and intercellular events and to direct new discoveries in biological and medical research.Hematopoietic stem cells (HSCs) form blood and immune cells and are responsible for the constant renewal of blood. To produce new blood cells, HSCs proliferate and differentiate to different blood cell types (1). To analyze stem-cell behavior and infer cell features, the localization, segmentation, and tracking of HSCs in culture is crucial. In our previous work we addressed cell detection/localization (2,3) and the association of detected cells (4). Yet to infer the cell features, we need to outline the boundaries of individual, touching, and dividing cells.Previously, in vitro time-lapse video microscopy has been used to identify phenotypic traits associated with in vivo HSC functionality. Specifically, longer cellcycle times were correlated with the retention of HSC activity, and the presence of lagging posterior projections (uropodia) was correlated with the loss of HSC activity (5).Various time-lapse studies were conducted on HSCs to provide information on their morphology, migration, and localization. For example, individual immature hematopoietic cells were examined to determine the differences in migration mechanisms caused by the primitive nature of the cells (6). This helped to explain the loss of phenotypic function during stem cell differentiation. Also a time-lapse video

Automatic Nuclei Segmentation in H&E Stained Breast Cancer Histopathology Images

“…Huang et al [14] described a method for segmentation of nuclei in hepatocellular carcinoma biopsy images based on marker-controlled watershed segmentation of initial contours followed by refinement with a snake model. Marker-controlled watershed, with markers produced by template matching, was also used by Kachouie et al [15] for segmentation of mammalian cells in microscopy images. For a broader overview of the topic of image analysis in histopathology images we refer the reader to a recent review [16].…”

Multi-Path Dilated Residual Network for Nuclei Segmentation and Detection

“…At present, active contour model [2], watershed model [3], and regional growth model [4] are three popular nuclear segmentation models. Among them, active contour model is used more frequently, mainly because it can fit the boundary of the target well.…”