Spot Edge Detection in cDNA Microarray Images using Window based Bi-Dimensional Empirical Mode Decomposition

Int. J. on Recent Trends in Engineering and Technology, Vol. 10, No. 1, Jan 2014

Spot Edge Detection in cDNA Microarray Images
using Window based Bi-Dimensional Empirical Mode
Decomposition
LakshmanaPhaneendra Maguluri1,M.N.Satish Kumar2, and V.jayaramakrishna3
1,2

Gudlavallegru Engineering College/CSE Department,Vijayawada,India
1
Email: phanendra51@gmail.com
2
Email:maganti.nagasatishkumar@gmail.com.
3
Andhra University/CSE Department, Vishakhapatnam, India
3
Email: v.jayaramakrinshna01@gmail.com

Abstract— Ongoing Microarray is an increasingly playing a crucial role applied in the field
of medical and biological operations. The initiator of Microarray technology is M. Schena et
al. [1] and from past few years microarrays have begun to be used in many fields such as
biomedicine, mostly on cancer and Diabetic, and medical diagnoses. A Deoxyribonucleic
Acid (DNA) microarray is a collection of microscopic DNA spots attached to a solid surface,
such as glass, plastic or silicon chip forming an array. Processing of DNA microarray image
analysis includes three tasks: gridding, segmentation and intensity extraction and at the
stage of processing, the irregularities of shape and spot position which leads to generate
significant errors. This article presents a new spot edge detection method using Window
based Bi-dimensional Empirical Mode Decomposition. On separating spots form the
background area and to decreases the probability of errors and gives more accurate
information about the states of spots we are proposing a spot edge detection via WBEMD.
By using this method we can identify the spots with low density, which leads to increasing
the performance of cDNA microarray images.
Index Terms— cDNA Microarray Image; Edge Detection; WBEMD; Image Processing

I. INTRODUCTION
Spot Edge Detection plays a fundamental role in processing of cDNA Microarray Images. While processing
microarray the edge includes very most important information, such that edges which are used for labeling
the points with a significant variation in gray level, and it can provide the position of object. Until now, we
have been applied some edge detections algorithms based on multi-resolution and gradient operators .Mean
while several gradient operators have Sobel[1], Prewitt[2], Marr[3]and Canny[4] are used. Multi-resolution
(multi-scaled) algorithms include wavelet based edge detection [5, 6] and multi-scaled mode based edge
detection [7, 8] have been used. Edge detection basically refers to identifying points of an image at which the
image brightness changes discontinuslly (sharply) The goal of Edge Detection is to finding sharp changes in
image brightness to capture the important changes in the properties of the events and to detect a desired
object , extract features of the image for further analysis.
An edge is the boundary between two different regions in an image. Edge detection is widely recognized as
the next revolution in image processing and computer vision. Edge detection can be broadly categorized into
DOI: 01.IJRTET.10.1.21
© Association of Computer Electronics and Electrical Engineers, 2014

two major approaches: search-based and zero-crossing based [1-2]. In search-based methods first compute
the strength of edge by using first-order derivative. While coming to the zero-crossing methods usually find
the edges by using second-order derivative of the image.
Microarray is an innovative tool in biomedical research which enables researchers to analyze genes, proteins
and other biological molecules. A microarray is a collection of spots containing DNA deposited on the solid
surface of glass slide. Each of the spot contains multiple copies of single DNA sequence [2]. Microarray
expression technology helps in testing and extracting the expression of large number of genes in parallel and
its applications can vary from cancer diagnosis to human identification. While preprocessing microarray
image is mainly to improve its defects including noise and poorly expressed spots. The processing of cDNA
Microarray Images constitutes of following steps: Gridding, Segmentation and Intensity Extraction. In order
to identify where exactly the spots are located, gridding is critical step in microarray image processing and it
is a process in which segments the microarray image into compartments Segmentation is the process of
dividing each compartment into spots and image background. In Intensity extraction, this calculates the
foreground and background intensity pairs. This Article is organized as follows: Section II presents Window
based Bi-Dimensional Empirical Mode Decomposition, Section III presents, Spot Edge Detection of cDNA
Microarray image, Section IV presents Experimental results, and finally Section V reports conclusion.
II. WINDOW BASED BI-DIMENSIONAL EMPIRICAL MODE DECOMPOSITION
In the upcoming part, the theory of WBEMD will be described; Firstly, the Bi-dimensional empirical mode
decomposition algorithm will be overviewed.
A. Bi-Dimensional Empirical Mode Decomposition overview
Bi-dimensional empirical mode decomposition is a 2D extension of the concept of empirical mode
decomposition, originally developed to analyze nonlinear and nonstationary 1D signals Besides BEMD, the
2D version of the EMD process is also known as 2DEMD,IEMDand so on. Processing of non-stationary and
nonlinear signal functions Huang et al. [5] proposed Empirical Mode Decomposition. The Empirical Mode
Decomposition differs from the traditional signal decomposition method which does not need to choose the
primary function in advance. Basically the empirical approach that can be applied to a data set, rather than a
theoretical tool. In this method any complicated data set can be decomposed into a finite and often a small
number of intrinsic mode functions and the decomposition are based on the local characteristic, it is
applicable to non-stationary processes. The Empirical approach has been successfully applied in and image
processing, medicine, fault diagnosis and so on. BEMD is applied to do the multi-scale decomposition to the
data. BEMD is a new, unsupervised, adaptive, proved reliable mono dimensional signals and sifting process
that decomposes a signal into its BIMFs and a bi-dimensional residue (BR) on the local frequency (or
oscillation information). Here instead of decomposition data or signal we are decomposing image, BEMD
decomposes an image into several bi-dimensional intrinsic mode functions (BIMFs) and a bi-dimensional
residue (BR). In Empirical Mode Decomposition has better time-frequency properties and it can be applied
in several fields like the mechanical fault detection, the electromagnetic wave and the acoustic signal
analysis. Now a day’s Empirical Mode Decomposition which grabs the attention of image analysts. The main
contribution of our approach is to apply the EMD to Spot Edge Detection in cDNA Microarray Images which
are widely recognized as a difficult and challenging problem in computer vision. IMFs satisfying the
following two conditions:
1. The number of extrema and zero-crossings must either equal or differ by at most one.
2. At any case, the mean value of the envelope defined by the local maxima and the local minima is
zero.
Bi-Dimensional Empirical Mode[10] Decomposition algorithm extracts Bi-dimensional Intrinsic Mode
Functions (BIMF) that is intrinsic to the original image through a sifting procedure. The decomposition
method in Empirical Mode Decomposition is called Shifting Process. The shifting process of the two
dimensional signals such as image can be adapted from the one-dimensional signal as follows.
1. Fix Ioriginal as a Microarray image to be decomposed. Let j=1 is the starting index of Intrinsic
Mode Function
I= Ioriginal
2. We have to spot the local maxima and local minima points in I.
3. Ccreate the upper envelope Eup of local maxima and the lower envelope Elw of local minima by
using interpolation approach.
24

4.
5.
6.
7.
8.

Work out the mean of the upper envelope and lower envelope.
Emean= [Eup + Elw]/2
Work out the Image Intrinsic Mode functions
Iimf = I- Emean
Repeat steps 2-5 until Iimf can be considered as an IMF
Recomputed Image Intrinsic values until jthIMF.
IMF (j) = Iimf, j=j+1, I = I- Iimf
Repeat steps 2-7 until all the standard deviation of two consecutive IMFs is less than a
predefined threshold.

B.Window based Bi-dimensional Empirical Mode Decomposition.
Empirical Mode Decomposition[9] analyzes the data not in the local area but in a whole area, which leads to
have gray spots, so that the high-frequency components having low-frequency components cannot be
separated. Here, the high-frequency component represents detail information and the low-frequency
component represents the trend information thus making a big splash in multi-scale signal analysis
I=∑

IMF + Res(1)

Where n is the required number of Intrinsic Mode Functions and Res Residue function. We can use squareadaptive sliding window or Windowed Fourier transform (WFT) in which the number of regional maxima is
equal to regional minima, is applied in WEMD to get the regional mean value then we have to work out the
intensity of the corresponding pixel in the transformed image and later on we have to calculate the first
IMFthat is to be subtracted from original image. This algorithm is listed as follows:
1. Fix Ioriginal as an Microarray image to be decomposed. Let j=1 is the starting index of Intrinsic
Mode Function,
I= Ioriginal
2. We have to spot the local maxima and local minima points in I.
3. Perform the following according to the value of IMF.
3.1) Set the max length of the quadrate window side N, where N must be odd and no less
than 3, so that the max area of the window is N*N;
3.2) Center the window at the first pixel (1,1) of I;
3.3) Let K=3, where K denotes the actual length of the quadrate window side;
3.4) Adjust the area of the window to K*K while keep the window centered at the current
pixel (m,n), and turn t (3.5)if K<N and the number of max in the window is not equal to
that of min, or else turn to (3.6);
3.5) Let K=K+2, then turn to (3.4);
3.6) Figure out c(m,n), which denotes the mean value of intensities of pixels in the window
centered at the pixel
(m,n)of r, and then turn to the next pixel and go back to (3.3) until all the
mean values is worked out;
4. Let C denote the matrix constructed by c(m,n), and let IMFi=I-C, I=C, and i=i+1;
5. .Let Res=I-IMFiand break if i reaches the required value, or else turn back to step (2)
WBEMD algorithm seems to be complex, but her actually need first Intrinsic Mode Function which contains
most quality Edge Information and counter equation can be calculated using following equation:
|
( , )|,
( , ) < 0
( , )=
− − − − − (2)
,
≥0
III. CDNA MICROARRAY SPOT EDGE DETECTION
A new spot edge detection of cDNA microarray image will be presented using Bi-dimensional Intrinsic
ModeFunctions (BIMF). The IMF that has been generated by applying empirical approach which contains
sufficient edge information and coming to next IMFs gradually reduced. Suppose if the contrast of the edge
description of the original microarray image is low, then the quality of the edges extracted by the first IMF
will be poor. If we want to improve the quality of the spot edges can be applied by using Bi Histogram
Equalization (BHE) [6] to the original Microarray image prior to the computation of WBEMD. Consider a
Microarray image with gray levels in the range [0, L-1] To [10]obtain the quality image following operations
should be meatWork out the probability density function (PDF)

25

Original Image

IMF 1

IMF 2

IMF 3

IMF 4

Fig 1: IMF of Microarray Image

( )=

= 0,1,2, … … . . , − 1 (3)

Calculate the cumulative distribution function (CDF)
C(r ) = ∑ p(r ) k = 0,1,2, … . . , L − 1 and o ≤ C( r ) ≤ 1
(4)
Where nkis the no of pixels in the image having gray level rkand rk is the kth gray level in the image.To
improve the contrast enhancement of microarray image we going for Bi Histogram Equalization to be
considered. It will find out the average gray level pixels in histogram of the microarray image[11] and then
sub divides the histogram in to two segments based on the pixels. Later on Bi histogram equalization
operation is applied to each segment.
The average of gray levels of the histogram is computed as
=∑

( )∗( )

(5)

Later on apply Window based Bi-Dimensional Empirical Mode Decomposition to the processed image to get
the first IMF. We can determine a suitable threshold to convert the gray level values of the IMF image into
black and white image such that the edge microarray image can be extracted. The processed edge image has
the thick edge lines and some interior isolated pixels. Then mathematical morphology is applied to remove
isolated pixels in the image. The final spot edge microarray image is obtained.
The main steps of the Spot Edge Detection using WBEMD are summed up as follows:
1. To obtain the sharper contrast of a microarray image, we are processing image with Bi Histogram
Equalization.
2. Then from obtained image apply WBEMD to get the first IMF.
3. To obtain the gray level values of the First IMF we have to choose suitable thresholds values T1 and
T2, which are helpful to convert the IMF image in to black and white image
( , )≤
, ≤
( , )=
(6)
1 ,
4. The processed Edge image has thick edge lines and isolated pixels.
5. Perform Mathematical Morphology operations on the resulting image to remove isolated pixels
6. Now we can obtain the ultimate spot edge image.
26

IV. EXPERIMENTAL RESULTS
The proposed method is used to detect the edge of cDNA microarray image. Processing of a cDNA
microarray image performed on a sample microarray slide that has 48 blocks, each block consisting of 110
spots. A sample chosen block has 36 spots of the block have been cropped for simplicity. The sample image
is a 198*196 pixel image that consists of a total of 38808 pixels. The microarray image is decomposed by
traditional EMD decomposition has the above mentioned defect, because the high-frequency components and
the low-frequency components can’t be effectively separated each other. WEMD can effectively deal with
the problem. The spot edge detection results of microarray image are presented in figure 2and the following
figure shows the speed of various EMD methods. Suitable thresholds of pixel values -75 and -42 were
determined through WBEMD, to convert into black and white image.
TABLE I: C OMPARING THE OPERATION SPEED OF THE VARIOUS E MD METHODS
METHOD
Time(Seconds)

BIEMD
62.639865

MTEMD
16.589472

DEMD
29.674407

WBEMD
0.286590

V. CONCLUSION
In this Article, we presented an new edge detection of cDNA Microarray images using WBEMD. The
WBEMD approach can separate the spots from the background and it will generate accurate information
about the spots such as pixel number and decreases the probability of errors encountered. The microarray
edge information obtained by using WBEMD is s better than Sobel and Prewitt operators. Experiments prove
that our approach is efficient and involves how to optimize the edges.
REFERENCES
[1] M.Schena, D.Shalon, Ronald W.davis and Patrick O.Brown,“Quantitative Monitoring of gene expression patterns
with a complementary DNA microarray”, Science, 270,199,pp:467-470.
[2] Wei-Bang Chen, Chengcui Zhang and Wen-Lin Liu, “An Automated Gridding and Segmentation method for
cDNAMicroarray Image Analysis”, 19th IEEE Symposium on Computer-Based Medical Systems.
[3] Tsung-Han Tsai Chein-Po Yang, Wei-ChiTsai, Pin-Hua Chen, “Error Reduction on Automatic Segmentation in
MicroarrayImage”, IEEE 2007.
[4] EleniZacharia and DimitirsMaroulis, “Microarray Image Analysis based on an Evolutionary Approach” 2008 IEEE.
[5] N.E.Huang, Z.Shen, S.R.Long, “The empirical mode decomposition and the Hilbert Spectrum for non-linear and
nonstationarytime series analysis”. Proc. Roy. Soc, London.A, Vol.454, pp.903-995, 1998.
[6] Ali Ziaei, HojatollahYeganeh, KarimFaez, SamanSargolzaei, “A Novel Approach for Contrast Enhancement in
Biomedical Images Based on Histogram Equalization”, 2008 International Conference on BioMedical Engineering
andInformatics.
[7] Jin Hyuk Kim, HyeYouns Kim and Yong Sung Lee, “ A novel method using edge detection for signal extraction
fromcDNA microarray image analysis”, Experimental and Molecular Medicine, Vol 33, No.2, 83-88, June 2001.
[8] Sharif M. A. Bhuiyan, Reza R. Adhami, Jesmin F. Khan,“Aurora Image Denoising with a Modified
BidimensionalEmpirical Mode Decomposition Method,” Proc. of IEEESoutheastCon 2008, pp. 527-532, 2008.
[9] S.Anand, N.M.MarySindhuja,” Spot Edge Detecton in Microarray Images Using Balanced GHM
Multiwavelet”,International Conference on Control, Automation, Communication and Energy Conservation, 2009
IEEE.
[10] James Z. Zhang “Edge Detection Using Fast Bi-dimensional Empirical Mode Decomposition and Mathematical
Morphology” in 978-1-4244-5853-0/10/$26.00 ©2010 IEEE
[11] Yao Pei, Yangang Wu, DachengJiaImage Denoising based on Bidimensional Empirical Mode Decomposition in
2011 International Conference on Mechatronic Science, Electric Engineering and ComputerAugust 19-22, 2011
IEEE.
[12] Zheng Ling Adaptive Multi-Scale Edge Detection in the EMD Domain in 2012 Fourth International Conference on
Computational and Information Sciences 978-0-7695-4789-3/12 $26.00 © 2012 IEEE
[13] J.Harikiran spot Edge Detection in Microarray Images using Bi-Dimensional Empirical Mode Decomposition
science Direct volume 4,2012,pages 227-231.c3it-2012

27

Spot Edge Detection in cDNA Microarray Images using Window based Bi-Dimensional Empirical Mode Decomposition

Recommended

More Related Content

What's hot (20)

Similar to Spot Edge Detection in cDNA Microarray Images using Window based Bi-Dimensional Empirical Mode Decomposition (20)

More from idescitation (20)

Recently uploaded (20)

Spot Edge Detection in cDNA Microarray Images using Window based Bi-Dimensional Empirical Mode Decomposition