Cell Cegmentation Using SEG-SELF and RFOVE methods

Introduction

Figure 1: A fluorescence microscopy image (left) and the output of the proposed method (right). The boundaries of the

detected cells and of the ground truth cell centroids are plotted with green curves and red pluses, respectively.

We present a novel solution to the problem of segmenting and splitting images of cells in an automatic and unsupervised manner [1] (SEG-SELF) and (RFOVE) [4].

Methodology of SEG-SELF [1]

Figure 2: (a) A fluorescence microscopy image. (b) The boundaries of the detected cells according to the Bradleys segmentation method projected on the given image. The cell centroids according to the ground truth data are plotted with red “+”. (c) The local backgrounds of the detected cells according to the Voronoi diagram of their centroids. The detected cells are plotted in black. (d) Final result of the SEG-SELF method..

  • The detection of cell nuclei is based on the Bradley’s method [2].
  • False positives are automatically identified and rejected based on shape and intensity features.
  • The proposed method is able to automatically detect and split touching cells.
  • To do so, we employ a variant of a region based multiellipse fitting method [3] that makes use of constraints on the area of the split cells.

Methodology of RFOVE [4]

Figure 3:(a) The green and blue points constitute the foreground. The figure shows the results of Decremental Ellipse Fitting Algorithm (DEFA) [3] when forced to estimate two ellipses. The two colors indicate the achieved assignment of points to ellipses. (b) DEFA result for the image of (a) when DEFA estimates a single ellipse. (c) The result from the application of the proposed RFOVE algorithm on the same data. The yellow points participate in the estimation on both ellipses. (d) The result of RFOVE in a synthetic image from [5]. (e) A fluorescence microscopy image and (f) the output of the proposed method. In (d) and (f) the boundaries of the detected objects are shown in green color and their ground truth centroids are plotted as red pluses.

  • RFOVE is completely unsupervised, operates without any assumption or prior knowledge on the object’s shape and extends and improves the Decremental Ellipse Fitting Algorithm (DEFA) [3].
  • Both RFOVE and DEFA solve the multi-ellipse fitting problem by performing model selection that is guided by the minimization of the Akaike Information Criterion on a suitably defined shape complexity measure.
  • In contrast to DEFA, RFOVE minimizes an objective function that allows for ellipses with higher degree of overlap and, thus, achieves better ellipse-based shape approximation.

Experiments - Downloads of SEG-SELF method [1]


Experiments - Downloads of RFOVE method [4]


and from https://1drv.ms/u/s!AljjSIq2AA2TujvNgi4WYWNiyiUq?e=5yB6Es [3,4,5]

    • You can download the experimental results of RFOVE presented in [4] (.rar).
    • See the corresponding readme.txt files for more details.

Related Publications

[1] C. Panagiotakis and A. Argyros, Cell Segmentation via Region-based Ellipse Fitting, IEEE International Conference on Image Processing, 2018.

[2] Derek Bradley and Gerhard Roth, “Adaptive thresholding using the integral image,” Journal of Graphics Tools,

vol. 12, no. 2, pp. 13–21, 2007. 2, 3, 4

[3] C. Panagiotakis and A. Argyros, Parameter-free Modelling of 2D Shapes with Ellipses, Pattern Recognition, vol. 53, pp. 259-275, 2016.

[4] C. Panagiotakis and A.A. Argyros, Region-based Fitting of Overlapping Ellipses and its Application to Cells Segmentation, Image and Vision Computing, Elsevier, vol. 93, pp. 103810, 2020.

[5] S. Zafari, T. Eerola, J. Sampo, H. Kälviäinen, H. Haario, Segmentation of overlapping elliptical objects in silhouette images, IEEE Trans. Image Process. 24 (12) (2015) 5942–5952.