Conference Paper

Level Set Segmentation of Cellular Images Based on Topological Dependence

DOI: 10.1007/978-3-540-89639-5_52 Conference: Advances in Visual Computing, 4th International Symposium, ISVC 2008, Las Vegas, NV, USA, December 1-3, 2008. Proceedings, Part I
Source: DBLP


Segmentation of cellular images presents a challenging task for computer vision, especially when the cells of irregular shapes
clump together. Level set methods can segment cells with irregular shapes when signal-to-noise ratio is low, however they
could not effectively segment cells that are clumping together. We perform topological analysis on the zero level sets to
enable effective segmentation of clumped cells. Geometrical shapes and intensities are important information for segmentation
of cells. We assimilated them in our approach and hence we are able to gain from the advantages of level sets while circumventing
its shortcoming. Validation on a data set of 4916 neural cells shows that our method is 93.3 ±0.6% accurate.

Download full-text


Available from: Weimiao Yu, May 26, 2014
  • Source
    • "Timings for all figures are presented in appendix B. In our future work, we would like to extend our statistical physics approach to include topological dependence [31]–[33] and the subspace Mumford–Shah model [9]–[11]. Eden clustering can be used to include a topological dependence constraint developed recently to segment crowded objects [31] [32]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Natural images are depicted in a computer as pixels on a square grid and neighboring pixels are generally highly correlated. This representation can be mapped naturally to a statistical physics framework on a square lattice. In this paper, we developed an effective use of statistical mechanics to solve the image segmentation problem, which is an outstanding problem in image processing. Our Monte Carlo method using several advanced techniques, including block-spin transformation, Eden clustering and simulated annealing, seeks the solution of the celebrated Mumford–Shah image segmentation model. In particular, the advantage of our method is prominent for the case of multiphase segmentation. Our results verify that statistical physics can be a very efficient approach for image processing.
    New Journal of Physics 01/2011; 13(2):023004. DOI:10.1088/1367-2630/13/2/023004 · 3.56 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Automatic and quantitative measurement of neurites is a challenging task, while it is critical in many neurological studies. We developed a fully automatic method to trace and quantitatively measure the neurites. Our measurements are validated by comparison with the semi-automatic NeuronJ and commercial software HCA- Vision. The results demonstrate that the measurements of the three approaches have no significant difference. We also apply our approach for a biological study on neurite outgrowth and the measurements of four different conditions are presented.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Fluorescent microscopy imaging is a popular and well-established method for biomedical research. However, the large number of images created in each research trial quickly eliminates the possibility of a manual annotation; thus, the need for automatic image annotation is quickly becoming an urgent need. Furthermore, the high clustering indexes and noise observed in these images contribute to a complex issue, which has attracted the attention of the scientific community. In this paper, we present a fully automated method for annotating fluorescent confocal microscopy images in highly complex conditions. The proposed method relies on a multi-layered segmentation and declustering process, which begins with an adaptive segmentation step using a two-level Otsu’s Method. The second layer is comprised of two probabilistic classifiers, responsible for determining how many components may constitute each segmented region. The first of these employs rule-based reasoning grounded on the decreasing harmonic pattern observed in the region area density function, while the second one consists of a Support Vector Machine trained with features derived from the log likelihood ratio function of Gaussian mixture models of each region. Our results indicate that the proposed method is able to perform the identification and annotation process on par with an expert human subject, thus presenting itself a viable alternative to the traditional manual approach.
    Artificial Intelligence Review 10/2013; 42(3). DOI:10.1007/s10462-013-9415-x · 2.11 Impact Factor
Show more