This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 186.217.236.157

This content was downloaded on 08/04/2015 at 20:08

Please note that terms and conditions apply.

Algorithms for automatic segmentation of bovine embryos produced in vitro

View the table of contents for this issue, or go to the journal homepage for more

2014 J. Phys.: Conf. Ser. 490 012125

(http://iopscience.iop.org/1742-6596/490/1/012125)

Home Search Collections Journals About Contact us My IOPscience

iopscience.iop.org/page/terms
http://iopscience.iop.org/1742-6596/490/1
http://iopscience.iop.org/1742-6596
http://iopscience.iop.org/
http://iopscience.iop.org/search
http://iopscience.iop.org/collections
http://iopscience.iop.org/journals
http://iopscience.iop.org/page/aboutioppublishing
http://iopscience.iop.org/contact
http://iopscience.iop.org/myiopscience


Algorithms for automatic segmentation of bovine

embryos produced in vitro

D H Melo 1, M Z Nascimento 2, D L Oliveira 3, L A Neves 4 and K
Annes 5

1,3 Center of Mathematics, Computing and Cognition, Federal University of ABC (UFABC),
Santo André, SP, Brazil
2 Faculty of Computing, Federal University of Uberlândia (UFU), Uberlândia, MG, Brazil
4 Institute of Biosciences, Letters and Science, Department of Computer Science and
Statistics, São Paulo State University (UNESP), São José do Rio Preto, SP, Brazil
5 Laboratory of Cellular and Molecular Biology of the Center of Natural Sciences and
Humanities, Federal University of ABC (UFABC), Santo André, SP, Brazil

E-mail: douglas.melo@ufabc.edu.br

Abstract. In vitro production has been employed in bovine embryos and quantification of
lipids is fundamental to understand the metabolism of these embryos. This paper presents a
unsupervised segmentation method for histological images of bovine embryos. In this method,
the anisotropic filter was used in the differents RGB components. After pre-processing step,
the thresholding technique based on maximum entropy was applied to separate lipid droplets
in the histological slides in different stages: early cleavage, morula and blastocyst. In the post-
processing step, false positives are removed using the connected components technique that
identify regions with excess of dye near pellucid zone. The proposed segmentation method was
applied in 30 histological images of bovine embryos. Experiments were performed with the
images and statistical measures of sensitivity, specificity and accuracy were calculated based on
reference images (gold standard). The value of accuracy of the proposed method was 96% with
standard deviation of 3%.

1. Introduction
In vitro production (IVP) has been employed to improved features genetics of bovine embryos
[1]. However, the increase in the levels of cytoplasmic lipid droplets in the IVP of bovine embryos
are relationship reduction of embryos after cryopreservation [2]. Therefore quantify the lipids is
fundamental to understand the metabolism of these embryos.

Some methods are applied to evaluate and measure the features of lipids in the investigation of
IVP of bovine embryos. Among the techniques applied, the Sudan black B staining is a method
that consists of a lipophilic dye, which reacts with lipids. Regions with lipids are represented by
darker colours compared to the rest of the embryo without lipid droplets. After Sudan black B
staining of bovine embryos, a microscope equipped with digital camera is used to acquire images
[3].

Some computational techniques have been applied to quantification of lipid droplets in images
of bovine embryos [3, 4]. However, the accuracy rate is low and these methods are supervised.
Thus, quantification of lipid by Sudan black B staining becomes one of the difficulties for the
specialists.

2nd International Conference on Mathematical Modeling in Physical Sciences 2013 IOP Publishing
Journal of Physics: Conference Series 490 (2014) 012125 doi:10.1088/1742-6596/490/1/012125

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Published under licence by IOP Publishing Ltd 1



This paper presents a unsupervised segmentation method for histological images of bovine
embryos. In this method, the anisotropic filter was used in the RGB colour components.
After pre-processing step, the thresholding technique based on maximum entropy was applied
to separate lipid droplets in the histological slides in different stages: early cleavage, morula
and blastocyst. In the post-processing step, false positives are removed using the connected
components technique that identify regions with excess of dye near pellucid zone. The proposed
segmentation method was applied in 30 histological images of bovine embryos.

2. Materials and Methods
The method was organized in three steps: (1) Pre-processing, (2) Segmentation and (3) Post-
Processing. Each step is described in detail in the following subsections.

2.1. Image Database
The database image used in the present study was made up of 30 histological slides in different
stages: early cleavage, morula and blastocyst. The images were acquired in the Laboratory
of Cellular and Molecular Biology of the Federal University of ABC. The microscopic images
were acquired using an inverted light microscope Olympus IX71 with digital camera coupled
(Lumenera Innity 1-1). Each slide was photographed at a magnification of 100x, using 10x
objective lens and ocular lens and image was saved in .png using the RGB model, 2048x1536
pixels, with 24 quantization bits.

2.2. Pre-processing
In the mounting process of each slice, fragments, dyes and other parts of the embryo can leak
out of the zona pellucida. The algorithm considers these problems by calculating the center of
the image and considering only the embryo centered. In this step, the pixels of the image were
compared with the coordinate of the center and the values were used to calculate the hypotenuse.
Regions with value 22% greater than the image area were removed.

Anisotropic diffusion filter can be used to smooth the outer regions of the edge giving a blur
effect and highlight the edges with a sharp effect keeping unchanged the inner regions of the
edges [5]. In this work, we applied this filter using the equation 1:

I(s, t+ 1) = I(s, t) +
λ

|ηs|
∑
p∈ηs

g(|OIs,p(t)|)OIs,p(t) (1)

where I(s, t) is the image, s is the pixel position, t is the number of interactions, ηs is the set
of neighbors space, λ is the diffusion speed, Is,p(t) is the gradient magnitude of the image I at
the point s in the direction (s, p) in iteration: tOIs,p(t) = I(p, t)− I(s, t), p ∈ ηs. In this paper,
we used parameter g defined by equation 2 .

g2(x) = exp

[
−x2

2σ2

]
. (2)

In equation 1, t = 20, λ = 0.2 and in equation 2, σ = 15.

2.3. Segmentation
In this step, a global thresholding was used, considering the maximum entropy technique.
This technique maximize entropy of the image dividing the histogram into two probability
distributions [6]. Each distribution represents one of the classes: background and object. The
entropy Hb(T ) and Hw(T ) associated to the pixels of the background and objects, respectively,
are expressed by

2nd International Conference on Mathematical Modeling in Physical Sciences 2013 IOP Publishing
Journal of Physics: Conference Series 490 (2014) 012125 doi:10.1088/1742-6596/490/1/012125

2



Hb(T ) = −
T∑
i=0

pi log pi (3)

Hw(T ) = −
L−1∑
i=T+1

pi log pi (4)

where ni is the number of pixels with gray scale i, pi is the probability of gray scale i found
in the image, n is the number of pixels of the image and L is the number of gray scales of image.

The gray scale is analyzed in order to find a threshold value T , such that T = Hb(T )+Hw(T )
is maximum. This technique defines a good separation between the object and the background
of the image, T = arg max[Hb(T ) +Hw(T )]

2.4. Post-Processing
The connected components technique was used to identify all segmented regions and remove
the regions that do not represent lipids: the 4-adjacency technique was used to remove regions
false-positives [7].

2.5. Quantitative Evaluation
The quantitative evaluation was performed by calculating the overlap between the image regions
segmented by the proposed method and the regions of a binary reference demarcated by a
specialist. This calculation was performed using statistical measurements of sensitivity (SE),
specificity (ES) and accuracy (ACC) ([8]).

3. Results and Discussion
The proposed segmentation method was applied to an image bank with 30 samples and measures
of sensitivity (SE), specificity (SP) and accuracy (ACC) were calculated. Figure 1 respectively
shows an image of embryo which was manually segmented by a specialist (gold standard) and
with the proposed method. Table 1 shows the measures considering sensitivity (SE), specificity
(SP) and accuracy (ACC).

The best ACC results (proportion of pixels defined as correctly segmented, both true positives
and true negatives) were obtained with the R colour component for the Early cleavage, Morula
and Blastocyst stages. Also, the proposed method provided segmented images with high rates
of SE and SP. In this study, sensitivity relates to the method’s ability to identify positive
areas or belong to a specific stage such as Early cleavage (SE of 94%), Morula (SE of 96% )
and Blastocyst (SE of 92%). These values were obtained with the B colour component. The
method’s ability to identify negative results (specificity) also is important: Early cleavage with
87%, Morula with 78% and Blastocyst with 88%, where these values were obtained with R
colour component. These measures suggest that textural features from bovine embryos images
are sufficient to separate lipids on different stages. Segmentation methods have been proposed
for investigation on different histological images but are not yet applied to bovine embryos.
This is a limitation for comparisons of our results. So, the proposal presented here provides a
significant contribution to studies focusing on IVP of bovine embryos.

4. Conclusion
This paper presented an unsupervised method for the automatic segmentation of bovine embryos
images considering different stages: early cleavage, morula and blastocyst. The proposed
method is effective for the segmentation of embryos images, considering the quantitative results
calculated. Also, the method offers the advantage of making automatically the segmentation.

2nd International Conference on Mathematical Modeling in Physical Sciences 2013 IOP Publishing
Journal of Physics: Conference Series 490 (2014) 012125 doi:10.1088/1742-6596/490/1/012125

3



(a) (b) (c)

Figure 1: Sample of an embryo is shown (a), the segmentation provided by a specialist is shown
in (b) and the result obtained with the proposed method is illustrated in (c).

Table 1: Measures of sensitivity (SE), specificity (SP) and accuracy (ACC) obtained for each
stage and colour component.

Stages Colour Channel ACC SE SP

Early cleavage Red 0.92 ± 0.04 0.92 ± 0.04 0.87 ± 0.12

Green 0.91 ± 0.03 0.91 ± 0.04 0.83 ± 0.22

Blue 0.93 ± 0.04 0.94 ± 0.04 0.68 ± 0.32

Morula Red 0.93 ± 0.04 0.93 ± 0.04 0.78 ± 0.30

Green 0.94 ± 0.04 0.95 ± 0.04 0.66 ± 0.33

Blue 0.96 ± 0.04 0.96 ± 0.04 0.56 ± 0.32

Blastocyst Red 0.92 ± 0.04 0.92 ± 0.04 0.88 ± 0.14

Green 0.92 ± 0.04 0.92 ± 0.04 0.84 ± 0.22

Blue 0.92 ± 0.04 0.92 ± 0.04 0.83 ± 0.24

In future studies, performance tests will be performed and the results compared to others
segmentation methods.

References
[1] Goovaerts I G F, Leroy J L M R, Langbeen A, Jorssen E P A, Bosmans E and Bols P E J 2012 Unravelling

the needs of singly in vitro-produced bovine embryos from cumulus cell co-culture to semi-defined, oil-free
culture conditions. Reprod Fertil Dev, 24 1084-1092.

[2] Mucci N, Aller J, Kaiser G, Hozbor F, Cabodevila J and Alberio R 2006 Effect of estrous cow serum during
bovine embryo culture on blastocyst development and cryotolerance after slow freezing or vitrification.
Theriogenology, 65 1551-1562.

[3] Sudano M, Paschoal D, Rascado T, Magalhaes L C, Crocomo L, de Lima-Neto J and Landim-Alvarenga F
2011 Lipid content and apoptosis of in vitro-produced bovine embryos as determinants of susceptibility to
vitrification. Theriogenology, 75 1211-1220

[4] Sudano M, Santos V, Tata A, Ferreira C, Paschoal D, Machado R, Buratini J, Eberlin M and Landim-Alvarenga
F 2012 Phosphatidylcholine and sphingomyelin profiles vary in bos taurus indicus and bos taurus taurus in
vitro- and in vivo-produced blastocysts. Biology of reproduction, 87 1-11.

[5] Perona P and Malik J 1990 Scale-space and edge detection using anisotropic diffusion. Pattern Analysis and
Machine Intelligence, IEEE Transactions on, 12 629-639.

[6] Pun T 1980 A new method for grey-level picture thresholding using the entropy of the histogram. Signal
Processing, 2 223-237.

[7] Gonzalez R and Woods R 2007 Digital Image Processing. Pearson.
[8] Sonka M 2000 Handbook of medical imaging: medical image processing and analysis, volume 2. Society of

Photo Optical.

2nd International Conference on Mathematical Modeling in Physical Sciences 2013 IOP Publishing
Journal of Physics: Conference Series 490 (2014) 012125 doi:10.1088/1742-6596/490/1/012125

4