Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
-
Upload
international-journal-for-scientific-research-and-development-ijsrd -
Category
Documents
-
view
215 -
download
0
Transcript of Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
7/26/2019 Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
http://slidepdf.com/reader/full/enhancement-and-segmentation-of-tumor-in-mammograms-for-breast-cancer-detection 1/3
I JSRD - I nternational Journal for Scientifi c Research & Development| Vol. 3, I ssue 12, 2016 | ISSN (onli ne): 2321-0613
All rights reserved by www.ijsrd.com 1039
Enhancement and Segmentation of Tumor in Mammograms for Breast
Cancer DetectionS.Savitha1 S.S.Thamilselvi2
1PG Student 2Assistant Professor1,2Department of Electronics & Communication Engineering
1,2
K.S.Rangasamy College of Technology, Tiruchengode, India Abstract — The well-known method used for the detection of
breast cancer is mammography. The mammogram images
are of poor contrast and noisy. To enhance the quality of the
mammograms the Robust polynomial filter is used in this
paper. It is the combination of Type-0 and Type-II
polynomial filter. The segmentation technique is then applied
for the separation of tumor region. The segmentation is
performed using Marker controlled watershed segmentation.
The pectoral muscles are removed from the image, the tumor
region is segmented. It helps to detect the breast cancereffectively.
Key words: Mammogram, Contrast Enhancement,
Polynomial Filter, Marker Controlled WatershedSegmentation
I. I NTRODUCTION
The main cause of death among women is due to breast
cancer. It is the second largest cancer in the world. It occursdue to either microcalcifications or masses. The
microcalcifications are tiny specks of calcium in the breast.
They are referred to as a cluster and may indicate a small
cancer. A mass is the group of cells clustered together more
densely than the surrounding tissue. The early detection of
the cancerous region helps in early diagnosis of a diseased person which will reduce mortality rates. For breast cancer
detection various techniques are used in which themammography is the mostly used technique by radiologist.
In mammography low dose X-ray is used. Due to this it is
difficult to interpret between normal and cancerous tissues.
The quality of the mammogram is improved using the imageenhancement technique. In some mammograms the
malignant tissues and pectoral muscles both are present. The
pectoral muscles are to be removed from the mammograms
for the effective detection of breast cancer. Since, the
pectoral muscles have same intensity and texture as that of
the malignant tissues, it is necessary to segment the pectoral
muscle before tumor detection. In this paper, for the
enhancement of mammograms Robust polynomial
filtering[4] technique is used. To segment the tumor regionthe Marker controlled watershed transform is applied [2].
II. METHODOLOGY
The input mammogram images are taken from MIAS and
DDSM databases. The images are normalized beforefurther processing. The Robust polynomial filter is applied
for the enhancement of mammograms. It is the combination
of Type-0 and Type-II polynomial filter. The Marker
controlled watershed segmentation is applied to the
enhanced results. The thresholding is then applied to remove
the pectoral muscles. The block diagram of the method is
shown in the Fig 1.
Fig. 1: Block diagram.
A. Robust Polynomial Filter
The Robust Polynomial Filtering[4] (RPF) is a linear
combination of Type-0 and Type-II polynomial filters. The
Type-0 filter provides contrast enhancement and noisefiltering and the Type-II filter provides edge sharpening.
The characteristic equation of the Robust Polynomial Filter
can be stated as:
y(n) = 0() + () (1)
Where:0() represents the Type-0 polynomial
filter and () represents the Type-II polynomial filter
respectively. The y(n) can be expanded as:
y(n) = + 0 +
(2)
Where:
= 052 + 2(1
2 + 32 + 7
2 + 92) +
2(22 + 4
2 + 62 + 8
2) (3)
0
= ∅052
+ ∅2(12
+ 32
+ 72
+ 92)
+∅2(22 + 4
2 + 62 + 8
2) (4)
= ∅7(1
3 + 1
7 + 3
9 + 7
9) +
∅8(19
+ 37
) + ∅9(28 + 4
6) + ∅10(1
6 +
18
+ 27 + 2
9 + 3
4 + 3
8 + 4
9 +
67) (5)
The terms a,b,c are the powers on the pixels raisedinside a 3×3 kernel: a for center pixel b for pixels at odd
locations and c for pixels at even locations.
B. Marker Controlled Watershed Segmentation
A marker is a connected component belonging to an image.
It possess the same intensity values and they are treated as
regional minima. Markers can be classified as foreground or background markers depending on its location of region of
interest. This gives a priori knowledge about segmentation.
Initially, the gradient magnitude of the input image is
computed.Next the morphological operations are applied to
mark the foreground. Following the opening with a closing
can remove the dark spots. The dark pixel in the image
belongs to the background. The gradient magnitude image is
modified so that it has only regional minima occur at
foreground and background. The watershed is then applied
to get the segmented result. The application of the watershed
transform results in over segmentation of the image due to
the presence of artifacts and noise. To avoid this, the
watersheds are applied to images with markers. For the
7/26/2019 Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
http://slidepdf.com/reader/full/enhancement-and-segmentation-of-tumor-in-mammograms-for-breast-cancer-detection 2/3
Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
(IJSRD/Vol. 3/Issue 12/2016/273)
All rights reserved by www.ijsrd.com 1040
visualization purpose the watershed output with markers is
converted to coloured images.
III. EXPERIMENTAL RESULTS AND DISCUSSION
The Fig 2.(a) and 2.(b) shows the mammogram images
taken from the two standard databases MIAS andDDSM.The normalizaton is performed before applying the
enhancement technique The optimal values of linear andquadratic filter coefficients determined for the Robust
Polynomial Filter [4] are: 0 = 0.2, 1 = 2 = 0.1, ∅0 = 8€,∅1
= ∅2= ∅6 = −€, ∅3 = −0.5€,∅4 = ∅5 = ∅10=€, ∅7= −2€, ∅8 =
−4€, ∅9 = 4,€ =0.15. The values of the weight indices are
taken as: a=8 μ, b=c=μ; the value of μ varies between 0.5 to
0.7 irrespective of the nature of the breast tissues.
The Fig. 3(a) and 3(b) shows the enhanced result
obtained using RPF technique. By using Robust PolynomialFiltering technique background tissues are suppressed and
the noises are reduced. It is shown that the tumor region and
pectoral muscles are present at the enhanced output.
Fig. 2(a): mdb184
Fig. 2(b): C_0080
Fig. 2: Original mammograms
Fig. 3(a): mdb184
Fig. 3(b): C_0080
Fig. 3: Enhanced results using RPFThe marker controlled watershed segmentation is
applied to the enhanced mammograms to segment the tumor
region. The Fig. 4 shows the results obtained using Marker
controlled watershed segmentation. The pectoral muscles,
tumor region and the background tissues are represented
with different colors.
The thresholding is then applied to the marker
controlled segmented output to remove the pectoral muscles.The Fig 5.(a) and 5.(b) shows the tumor region segmented
from the mammograms. The results shows only the tumor
region and the other regions are eliminated.
Fig. 4(a): mdb184
Fig. 4(b): C_0080
Fig. 4: Results obtained using the marker controlledwatershed segmentation.
Fig. 5(a): mdb184
7/26/2019 Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
http://slidepdf.com/reader/full/enhancement-and-segmentation-of-tumor-in-mammograms-for-breast-cancer-detection 3/3
Enhancement and Segmentation of Tumor in Mammograms for Breast Cancer Detection
(IJSRD/Vol. 3/Issue 12/2016/273)
All rights reserved by www.ijsrd.com 1041
Fig. 5(b): C_0080
Fig. 5: Results obtained for segmentation of tumor from
mammogram images.
Fig. 6(a): mdb241
Fig. 6(b): mdb315
Fig. 6: Results obtained for segmentation of tumor frommammogram images.
IV. CONCLUSION
The enhancement using Robust polynomial filter removes
the background noise and ill effects effectively. From the
segmented results it is proved that the presented work
effectively removes the pectoral muscles and segment out
only the tumor region. This helps to do feature extraction
and better classification of tumor region for the successful
detection of breast cancer.
R EFERENCES
[1] Yasmeen Mourice George, et al., “Remote computer -
aided breast cancer detection and diagnosis system
based on cytological images”, IEEE Systems Journal,
vol. 8, no. 3, pp.949-964, Sep. 2014.
[2] Ravi S and A M Khan, “Bio-medical imagesegmentation using marker controlled watershed
algorithm: a case study” International Journal of
Research in Engineering and Technology, vol.03,
pp.26-30, May 2014.
[3] Xiaoming Liu and Jinshan Tang, “Mass classification in
mammograms using selected geometry and texturefeatures, and a new SVM-based feature selection
method”, IEEE systems journal, pp.1-11, Nov. 2012.
[4] V.Bhateja,et al., “A Robust polynomial filtering
framework for mammographic image enhancement
from biomedical sensors,” IEEE Sensors Journal vol.
13,no.11, pp. 4147 – 4156, Nov. 2013.[5] Pandey,et al., “Design of new volterra filter for
mammogram enhancement,” in Proc. Int. Conf.
FICTA, vol. 199.Dec. 2012, pp. 143 – 151.
[6] S.Singh and K.Bovis, "An Evaluation of Contrast
Enhancement Techniques for Mammographic Breast
Masses," IEEE Transactions on Information
Technology in Biomedicine, vol. 9, no.1, pp.109-119,
2005