JOURNAL OF INFORMATION, KNOWLEDGE AND RESEARCH IN
BIOMEDICAL ENGINEERING
SEGMENTATION AND ANALYSIS OF
MICROSCOPIC AND MICROARRAY BONE IMAGES
ANAND JATTI
Assistant Professor, (Research Scholar, VTU, Belgaum), Instrumentation Technology
Department, R.V.College of Engineering, Bangalore, Karnataka, India
anandjatti@yahoo.com
ABSTRACT : Analysis of human bone cross sections represents a major area of research in bone biology. The
normal bone and bone cancer ‘Osteosarcoma’ case has been taken up for the current study, since it is the most
frequent type of bone tumor and is most common between the ages of 15 to 25. One of the biggest challenges
faced in image analysis is at the microscopic cellular level of bone and also at its gene level. In this paper
analysis of microarray and microscopic images of normal human and osteosarcoma has been done. Using
microarrays, expression levels of thousands of genes can be measured simultaneously. In Microarray
experiment, the labeled nucleic acids are derived from the miRNA of a bone sample and so the microarray
measures the gene expression. From this technique, the obtained microarray image consists of genes
responsible for osteosarcoma case, by quantification of such genes of bone cell, able us to know the
comparative study between the normal and abnormal case of human bone tissue. In biomedical research, the
revolution of microscopy imaging of large amount of image data renders manual visual analysis impossible,
requiring automated image analysis systems. The qualitative analysis of digital microscopic images provides
better visualization to human, becomes the predominant. And the quantitative information of such images
pertains to morphology of the object (tumor) under study or its function, therefore extraction of such
information from images is the main issue to be tackled.
KEY WORDS:- Bone Cancer Osteosarcoma, Image Processing, Microscopic Image Analysis, Micro Array
Image Analysis.
1. INTRODUCTION
The Human bone cancer case ‘Osteosarcoma’ also
called as Osteogenic sarcoma is a malignant primitive
bone tumor, usually developing in children and
adolescents, with a high tendency to metastasize.
Metastasis is the dissemination of cancer cells from
the primary tumor to a distant organ and represents
the most frequent cause of death for patients with
cancer. This project work has two implications, First,
at gene level analysis of human bone tissue of both
normal and osteosarcoma case is through microarray
technique, in which the microarray images were
produced through a series of steps and procedure at
genetic laboratory and it is analyzed for intensity
extraction of each miRNA detected spot in the image
which corresponds to single gene expression. The
intensity of miRNA can be compared in both the
normal and abnormal bone tissue.
The microRNA are single stranded RNA molecules
of about 21 to 23 nucleotides in length which
regulates gene expression, and these are transcribed
from DNA but not translated into proteins. miRNA
microarray is a high throughput method for studying
microRNA expressions in cultured cells or tissues
which facilitates the study of biological roles of
microRNA.
A Microarray consists of a grid of tiny spots of DNA
printed on a surface, with each spot usually
corresponding to a different gene. miRNA, reflecting
which genes are currently active, is extracted from
cells and processed to form single stranded nucleic
acids labeled with fluorescent molecules. The labeled
nucleic acids bind to their corresponding microarray
spot. The microarray is then scanned with a laser
confocal microscope (Agilent Scanner).
The Microscopic images are obtained through the
snap shot of microscopic picture on slide by a digital
camera.
The importance of Image processing of microarray
and microscopic are as follows: The analysis of
digital microarray and microscopic images of human
bone samples obtain qualitative and quantitative
results. Microarray qualitative includes Gridding,
spot finding and background correction. Microarray
quantitative analysis was useful in providing the
intensity levels of each miRNA detected spots which
correspond to the single gene expression.
Microscopic qualitative analysis gives detailed
information about the processing of the foreground
and background of the digital images which
contained cancerous cells. The microscopic
ISSN: 0975 – 6752 | NOV 09 TO OCT 10 | Volume 1, Issue 1
Page 1
JOURNAL OF INFORMATION, KNOWLEDGE AND RESEARCH IN
BIOMEDICAL ENGINEERING
quantitative analysis provides the information about
the area of affected cells in a given digital image.
2. MATERIALS AND METHODOLOGY
A. Materials used for Microscopic Images: The
glass slides of the biopsy samples are placed below
the digital camera focusing point and the digital
microscopic pictures were taken for further analysis.
B. Materials used for Microarray Images: The
normal and osteosarcoma of human bone sample
tissue shown in Fig.1 was obtained in the form of
FFPE from the Institute of Oncology. The FFPE
tissues have been widely used to archive samples
obtained from biopsies of human bone cancers and
diseased tissues like linked to genetic abnormalities.
The FFPE samples are then given for the RNA
extraction for further process. The following are the
samples for the analysis:
Normal Bone
Osteosarcoma
Fig.1 FFPE samples of Normal bone and
Osteosarcoma tissue.
Methodology
Methodology for Microscopic Image Analysis:
The procedure and techniques carried out for bone
image analysis is shown in flow chart.
Fig.2 Work flow chart.
Image acquisition: Image acquisition is the first
process involved. The microscopic bone cross section
image acquired by using electronic microscope.
Biological aspect of bone: Important features in the
bone cross section such as harvesian canals, osteons,
osteon fragments, lamellar bone, bony trabeculae,
myxoid matrix and artifact for different age groups
and also for age related developments are observed.
Image format conversion: The digital image
obtained using electronic microscope is in RGB
(Red, Green and Blue) format and converted to grey
level image for further processing. The MATLAB
tool is used for image format conversion from RGB
to gray level.
Pre processing of an image:
The preprocessing operation is carried out to extract
details that are obscured in an image or to highlight
and classify certain features of an interest in an
image.
The tumor cells in a digital microscopic image is
analyzed and processed by using MATLAB Image
processing toolbox. Microscopic bone images were
analyzed through following steps:

The morphological operations are applied to
image, for the extraction of the image components.

The Watershed segmentation technique used
to enhance the tumor cells in the image.

The average area of tumor is then calculated
with respect to the total image area in terms of pixels
and tabulated results.
Methodology for Microarray Image Analysis: The
importance of Image processing of DNA microarray
is as follows: The analysis of digital DNA microarray
images of human bone samples obtains qualitative
and quantitative results. DNA Microarray qualitative
image analysis includes gridding, spot finding and
background correction [4]. DNA Microarray
quantitative image analysis was useful in providing
the intensity levels of each miRNA detected spots
which correspond to the single gene expression. The
steps of DNA Micro array Image formation from
bone samples are as follows

RNA extraction from FFPE bone samples.

Labeling and Hybridization - Control and
test miRNA samples are processed on the same array
using Cyanine-3 fluorophore.

Scanning and feature extraction.

miRNA profile is obtained.

The obtained DNA Microarray image for
both normal and osteosarcoma human bone samples
were analyzed through Gene Spring software.
3. RESULTS
Results of Microscopic Image Processing:
The tumor cells in a digital microscopic image is
analyzed and processed by using MATLAB Image
processing toolbox. Microscopic bone images were
analyzed through following steps:
ISSN: 0975 – 6752 | NOV 09 TO OCT 10 | Volume 1, Issue 1
Page 2
JOURNAL OF INFORMATION, KNOWLEDGE AND RESEARCH IN
BIOMEDICAL ENGINEERING

The morphological operations are applied to
image, to extract the image components.

The Watershed segmentation technique used
to enhance the tumor cells in the image.

The average area of tumor are then
calculated with respect to the total image area in
terms of pixels and tabulated for total of twelve
images.
The image processing algorithm was applied to
digital images. Qualitative analysis includes
morphological operations such as background
suppression, enhancement of contrast and watershed
segmentation. Quantitative analysis includes the
measurement of area of tumors in the image by
segmentation of each tumor pixel. The result of one
tumor image is presented in this paper. The input
image can be seen in Fig.3 (a). The results for these
images are shown correspondingly.
Morphological
Operation:
To
perform
morphological operations on an image, the image has
to be in binary format, thus input image is converted
to binary image as shown in Fig.3 (a), (b) and (c).
(a)
(b)
(c)
Fig.4
(a) Imposed minima
(b) Watershed (c) Labeled regions
Measurement of Area: The average area of the
tumor or cancerous cells in an image is found and
calculated with respect to the total area of the input
image in terms of pixels. Following steps are
performed to obtain the quantitative measurement of
microscopic image. First the Histogram plot of gray
scale image has been taken. Histogram is a
construction of the counts of pixels with particular
amplitude value as shown in Fig.5 (a). It enhances the
contrast of the original image through equal intensity
distribution. The enhanced image is converted to
binary image by fixing a threshold either 0(black) or
white (1) as shown in Fig.5 (b).
(a)
(b)
(c)
Fig.3 (a) Original Image. (b) Binary Image
Dilation
into any watershed region are given a pixel value of
zero. View the label matrix as an image using
“label2rgb’ as shown in Fig.4(c). RGB color image is
for the purpose of visualizing labeled regions.
c)
Watershed Segmentation Technique:
Fig.4 (a) detects all intensity valleys below a
particular threshold with the ‘imextendedmin’
function. The output of the ‘imextendedmin’ is a
binary image. The location rather than the size of the
regions in the ‘imextendedmin’ image is important.
The “watershed” function returns a label matrix
containing non negative numbers that correspond to
watershed regions in Fig.4 (b).Pixels that do not fall
(a)
(b)
ISSN: 0975 – 6752 | NOV 09 TO OCT 10 | Volume 1, Issue 1
Page 3
JOURNAL OF INFORMATION, KNOWLEDGE AND RESEARCH IN
BIOMEDICAL ENGINEERING
(c)
Fig.5 (a) Histogram plot. (b) Threshold (c) Labeled
regions
Individual’s grains are labeled with different colors
as shown in Fig.5(c) and extracted area gray level
values from detected objects can be seen in Fig.5 (a).
Looking at the edges of the detected grain by finding
the perimeter pixels of the image using ‘bwperim’.
Different colors for different perimeter pixels are
assigned as shown in Fig.5(b). Background is
suppressed by using the subtraction method.
(b)
Fig.7 (a) Comparison of size distribution.
(b) Final detected grains.
The summation of area of detected grains or tumor is
found out with respect to the total area of the image
in terms of pixels. The table shows the total area (in
pixel) of the image and affected area (average tumor
area), and area unaffected (without tumor).
Table 1: Sample case study of Bone Normal and
Abnormal samples (Osteosarcoma).
Image
Osteo1
(a)
Osteo2
Norma
l1
Norma
l2
b)
(c)
Fig.6 Segmentation without background. (b) New
Size distribution. (c) Bordered objects removed.
Total
Area
(in
pixel)
30000
00
30000
00
56250
0
56250
0
Area
affecte
d
(in
pixel)
147513
32855
Area
unaffec
ted
(in
pixel)
285248
7
287660
0
529645
14276
548224
123400
Results from DNA Micro Array Technique:
The images obtained after scanning are subjected to
image analysis using the FE. Later these are analyzed
by Gene Spring GX 10, for further analysis. The
sample given was one normal and one osteosarcoma
bone FFPE tissue, the replication of each sample was
done for the comparison purpose. Thus generated
total four images, two from control sample (Normal
tissue), and two from non control sample (abnormal
or osteosarcoma tissue).Microarray image of normal
and osteosarcoma case can be seen in Fig.8.
Segmentation is done once more after image
subtraction to get clear distinction between detected
objects. New areas then extracted after segmentation
can be seen in Fig.6 (b). Comparison of original and
background size distribution is done. Removal of
partial grain cut off by edges to get clear distinction
can be seen in Fig.6(c).
Fig.8 DNA Microarray Image. (a) Normal or control.
(b) osteosarcoma or Non control.
(a)
The data analysis is done by using the software
GeneSpring, version GX 10. Provides powerful,
accessible statistical tools for fast visualization and
analysis of expression data. Designed specially for
ISSN: 0975 – 6752 | NOV 09 TO OCT 10 | Volume 1, Issue 1
Page 4
JOURNAL OF INFORMATION, KNOWLEDGE AND RESEARCH IN
BIOMEDICAL ENGINEERING
the needs of biologists. GeneSpring GX offers an
interactive desktop computing environment that
promotes investigation and enables understanding of
DNAmicro array data within a biological context.
The obtained intensity values of miRNA detected in
normal and abnormal were tabulated and generated
cluster, scatter plots.
The intensities and background subtracted were taken
from the raw data file generated from feature
extraction. The obtained values and image from the
feature extraction is further subjected to data analysis
by the use of GeneSpring GX 10.The sum of
background subtracted signals for each repeated
miRNA was calculated. To compare between the
samples further normalization (50th percentile
normalization) step was performed. (50th percentile
normalization: values below 0.01 were set 0.01.Each
measurement was divided by the 50th percentile of
all measurements).After experiment, the total number
of
probes
in
array
excluding
controls
is:11180.Number of miRNA’s repeated more than
one time with different probe ids:534.The repeated
miRNA were grouped in both normal and cancer
bone, as shown in the Table.1.
Table 2. miRNA detection in cancer and normal
tissue.
Group
miRNA
Detected
Cancer tissue compared to
normal bone
32
Normal bone compared to
cancer tissue
75
The ‘Fold change’ cutoffs have been widely used in
microarray assays to identify genes that are
differentially expressed between query (cancerous)
and references (normal) samples. More accurate
measures of differential expression and data
normalization strategies are required to identify high
confidence sets of genes with biologically
meaningful changes in transcription. The results are
obtained and tabulated, then plotted for the same. The
Heat map was generated for the arrays hybridized
based on summed background-subtracted signals of
each repeated miRNA’s. The overview of all the
miRNA’s in the array is as shown in the form of
cluster in the Fig.9.
Fig.9 Overview of all the miRNA’s in an array.
Yellow: Non detected miRNA’s. Green: miRNA with
low detection level. Red: miRNA with low detection
level.
Scatter plot showing the miRNA’s distributed based
on their expression values miRNA’s in Red blocks
indicate the miRNA’s detected. These are shown in
scatter plot can be seen in Fig.10.
Fig.10 Scatter plot showing the miRNA’s distributed
based on their expression values miRNA’s in Red
Blocks indicates the miRNA’s detected in cancer
bone (fold>1) compared to normal bone (fold<0.5).
All the red blocks fall under positive values. Note: X
axis: Cancer bone and Y axis: Normal bone.
Line Plot: Line plot is created for the values of both
normal and cancerous bone and also only for cancer
bone can be seen in Fig.11.The plot provides the
comparison between normal and abnormal tissues.
Plot in blue color indicates the cancer and red plot
indicates the normal fold values.
5
Fol d_Canc Ti s s ue
hs a-mi R-24
Fol d_N Bone
4
hs a-mi R-33b
hs a-mi R-15b
3
hs a-mi R-143
hs a-mi R-107
hs a-mi R-26a
hs a-mi R-452*
hs a-mi R-320
hs a-mi R-422b
hs a-mi R-30c
hs a-mi R-433
hs a-mi R-130a
hs a-mi R-103
hs a-mi R-203
hs2a-mi R-101
hs a-mi R-152
hs a-mi R-106b
hs a-mi R-138
hs a-mi R-16
hs a-mi R-200b
hs a-mi R-330
hs a-mi R-30e-5p
1
hs a-mi R-629
hs a-mi R-496
hs a-mi R-495
hs a-mi R-564
hs a-mi R-568
hs a-mi R-616
hs a-mi R-641
hs a-mi R-615
hs a-mi R-487a
hs a-mi R-520d*
hs
ahs miR
a
hs -m -10
a- iR 1
m hs iR 103
a hs -m 106
a- iR b
m hs iR 107
a- -1
hs miR 30a
a- hs miR 138
a hs -miR 143
a- m 1
h iR 52
hs sa- -15
a- miR b
m
hs iR -16
a- -2
m 0
hs iR 0b
a hs -m 203
a- iR
h m -2
hs sa iR- 4
a- -m 26
m iR a
iR hs -3 30
a- 0e c
hs miR -5
a- - p
hs miR 320
a
hs -m -33
a- iR 0
m hs iR 33b
a hs -m 422
a- iR b
hs miR -43
a- - 3
m 4
hs iR 52*
a- -4
h m 87
hs sa- iR-4 a
a- miR 9
m - 5
hs iR-5 496
a- 2
hs miR 0d*
a- hs miR 564
a- hs miR 568
a- hs miR 615
a- hs miR 616
a- m 62
iR 9
-6
41
0
-1
-2
(a)
ISSN: 0975 – 6752 | NOV 09 TO OCT 10 | Volume 1, Issue 1
Page 5
JOURNAL OF INFORMATION, KNOWLEDGE AND RESEARCH IN
BIOMEDICAL ENGINEERING
(b)
Fig.11 Line plot. (a) Expression level of detected
miRNA’s in Normal bone. (b) Line graph of cancer
bone. All the miRNA’s expression level are >1. The
graph represents the fold values of the detected
miRNA’s in cancer bone. X-axis-miRNA’s, Y-axis:
Fold value.
4. DISCUSSION AND CONCLUSION
The work presented in this project shows some very
interesting results in a field where very little research
has been done on bone image analysis. The Human
Bone samples are obtained from the Institute of
Oncology, of both normal and osteosarcoma case has
been taken up and they are analyzed at genetic level
(microarray) and microscopic cellular level.
The Emerging Microarray Technology has been
utilized to get the microarray image of human bone
samples for both normal and osteosarcoma case, and
intensity of each detected miRNA in both normal and
osteosarcoma has been tabulated. The miRNA
quantification is achieved from this powerful
technique.
The microarray image analysis of bone in MATLAB
provides a way to extract the intensity of gene
expression or which corresponds to each spot in the
image. The microscopic images are analyzed and the
area of tumor in the images are obtained and
tabulated in the results. Dark spots indicate tumors in
all the three cancerous images are segmented and the
area of each spot is calculated. This is useful to know
the area of tumor present and can be diagnosed
accordingly.
5. FUTURE ENHANCEMENTS
The detected miRNA expression for the human bone
sample from the microarray technique, are associated
with clinical variables of cancers, so the miRNA’s
can be readily used for cancer diagnosis and
prognosis. Each detected miRNA in cancer can be
studied for further research and drug development.
The major roles of miRNA’s in regulating genetic
regulatory circuit and also in the response of immune
system give us a clue of the impact miRNA would
have on Vaccine research and development.
6. REFERENCES
[1] Luis Hernandez, Plula Gothreaux & Liwen Shih,
“Toward Real-Time Biopsy image analysis & Cell
segmentation”, Computer Engineering, University of
Houston, 2004.
[2] Zhi-Qiang Liu, Hui Lee Liew, John G.Clement
and
C.David
L.Thomas,
“Bone
Image
Segmentation”, IEEE Transactions on Biomedical
Engineering. Vol.46, No.5, May 1999.
[3] Zhi-Qiang Liu, Hui Lee Liew & Sandy Dance,
“Image processing techniques for Quantitative Bone
Image Analysis”, International Symposium on Signal
processing & its applications, IEEE/ISSPA, August
1996.
[4] Zhi Qiang Liu, Timpthy J. Austin & Daniel
Moore, “Image processing Techniques for Bone
image analysis”, Computer Vision & machine
intelligence lab, IEEE 1995.
[5] Young Sun Lee and Anindya Dutta, “MicroRNA
in Cancer”, Annual Review of pathology, Mechanism
of disease journal, 2009.
[6] Md. Maruf Monwar of Siamak De Zai, “A
parallel processing approach for performation
Extraction from Microarray images”, IEEE, May
2006
[7] Basim Alhaddi, Hussam Nawwaf Fakhouri &
Omar S.AlMousa, “DNA Microarray Genome Image
processing using fixed spot position”, American
journal of applied Sciences 3(2): 1730-1734, 2006.
[8] Robert S.H. Istepanian, “Microarray image
processing: Current status & future Directions” IEEE
Transactions on NanoBioScience, Vol.2, No.4,
December 2003.
[9] R.C.Y. Cheung, Christopher J.S.Desilva,
“Analysis of Gene Microarray Image”, centre for
intelligent information processing Systems, IEEE
1999.
7. ACKNOWLEDGEMENT
I acknowledge for their support for the smooth
completion of the project work to the Department of
Science and Technology (DST), Government of
India, New Delhi for giving financial support for
carrying out this project work.
ISSN: 0975 – 6752 | NOV 09 TO OCT 10 | Volume 1, Issue 1
Page 6