Summary
Breast cancer is the most common type of cancer in women, which also causes the most cancer deaths among them today. Mammography is the only reliable method to detect breast cancer in the early stage among all diagnostic methods available currently. Breast cancer can occur in both men and women and is defined as an abnormal growth of cells in the breast that multiply uncontrollably. The main factors which cause breast cancer are either hormonal or genetic. Masses are quite subtle, and have many shapes such as circumscribed, speculated or ill-defined. These tumors can be either benign or malignant.
Computer-aided methods are powerful tools to assist the medical staff in hospitals and lead to better and more accurate diagnosis. The main objective of this research is to develop a Computer Aided Diagnosis (CAD) system for finding the tumors in the mammographic images and classifying the tumors as benign or malignant. There are five main phases involved in the proposed CAD system: image pre-processing, extraction of features from mammographic images using Gabor Wavelet and Discrete Wavelet Transform (DWT), dimensionality reduction using Principal Component Analysis (PCA) and classification using Support Vector Machine (SVM) classifier.
Computer-aided methods are powerful tools to assist the medical staff in hospitals and lead to better and more accurate diagnosis. The main objective of this research is to develop a Computer Aided Diagnosis (CAD) system for finding the tumors in the mammographic images and classifying the tumors as benign or malignant. There are five main phases involved in the proposed CAD system: image pre-processing, extraction of features from mammographic images using Gabor Wavelet and Discrete Wavelet Transform (DWT), dimensionality reduction using Principal Component Analysis (PCA) and classification using Support Vector Machine (SVM) classifier.
Excerpt
Table Of Contents
5
CHAPTER 1 : INTRODUCTION
1.1 Introduction
Breast cancer represents the most common cause of cancer deaths in women today and it is the
most common type of cancer in women. Mammography is the only reliable method which is
used to detect breast cancer in the early stage among all diagnostic methods available currently.
Breast cancer is defined as an abnormal growth of cells in the breast that multiply uncontrollably.
The main factors which cause breast cancer are either hormonal or genetic. Breast cancer can
occur in both men and women. Masses are quite subtle, and have many shapes such as
circumscribed, speculated or ill-defined. Tumors can be either benign or malignant.
A benign tumor is not cancerous because:
1. Benign tumors do not invade healthy surrounding tissue
2. They do not grow and if removed normally don't grow back.
3. Benign tumors do not spread to other parts of the body i.e. metasize.
A malignant tumor is cancerous because:
1. Malignant tumor cells can invade and damage surrounding tissue.
2. If the tumor is removed it can grow back.
3. Malignant tumor cells can meta-size.
Natural images typically contain distinctive and regular patterns within their spatial
texture. Those specific patterns would be emphasized while others, sought as noise in this
context, should be flattened or even discarded. Examples of specific patterns include edges,
corners and interest (feature) points. This process is similar to feature detection, which is
typically achieved employing filtering in spatial or frequency domain. X-ray imaging of the
breast also known as screening mammogram is the most effective tool for early detection of
breast cancer. Radiologists may visually search mammograms for the detection of abnormalities.
Early diagnosis and screening is crucial for successful medical treatment or cure. Artifacts
appearing in the mammogram images could indicate a potential presence of a benign or
malignant tumor. Important visual clue of breast cancer can be calcification clusters or
preliminary signs of masses.
Calcium deposits and Masses can be easily identified by visual inspection in X-ray
images as they are much denser than all other types of surrounding soft tissues. Unusual smaller
6
and clustered calcifications are associated with malignancy while there are other calcifications
like diffuse, regional, segmental and linear are typically benign. Such calcifications are known as
micro-calcifications. Automatic tumor detection is very challenging as the suspicious masses
may appear as free shape and irregular texture, so no precise patterns can be associated with
them. Breast tumors usually appear in the form of dense regions in mammograms. A benign
mass has a round, smooth and well circumscribed boundary and a malignant tumor usually has a
speculated, rough, and blur boundary. This system is developed for analysis of digital
mammograms using Gabor Wavelet and Discrete Wavelet Transform (DWT). The proposed
system has phases like preprocessing the image, Feature Extraction from the preprocessed image
and Classification of mammogram as Benign or Malign. The Image Preprocessing phase
involves image acquisition and enhancement of image. Preprocessing is needed to improve the
quality of the images and make the feature extraction phase more reliable.
Features of the objects carefully are representative of the most relevant information of the
image, if selected accurately may offer a complete characterization. Feature extraction
methodologies analyze images to extract the most prominent features that can be used for
classification of objects into various classes. Here Gabor Wavelet based features are used. Once
features are extracted, Principal Component Analysis (PCA) is applied to it for dimensionality
reduction. Finally, the extracted features are passed to the Support Vector Machine Classifier and
are classified into normal or abnormal (benign or malignant) images. For comparison Discrete
Wavelet Transform (DWT) is applied. The proposed system is applied to 322 mammogram
images, originating from the MIAS database. The results are analyzed using MATLAB.
1.2 Necessity
Breast cancer is the most common cause of cancer in women. The chance of a woman having
breast cancer during her life is about 1 in 8. The chance of dying from breast cancer is about 1 in
35. Large numbers of mammograms are generated daily in hospitals and health checkup centers.
Thus, Radiologists Physicians have more and more images to analyze manually. After analyzing
a number of images, the process of diagnosing them becomes more susceptible to errors. Thus,
computer-aided diagnosis (CAD) system may be used to assist the physician's work to reduce
mistakes. Thus, Developing CAD systems to be used in medical care is becoming highly
important, and this helps the radiologists use the result as a "second opinion" to assist them for
7
speeding up the diagnosing task. Mammography is one of the most reliable methods for breast
cancer detection. Currently, X-ray mammogram is considered a standard procedure for breast
cancer diagnosis. However, retrospective studies have shown that radiologists can miss the
detection of a significant proportion of abnormalities.
Computer-Aided Detection (CAD) systems may be used to aid radiologists in detecting
mammographic lesions that may indicate the presence of breast cancer. These systems may act
only as a second reader and the final decision is made by radiologist. Recent studies have also
shown that Computer Aided Detection (CAD) systems have improved accuracy of detection of
breast cancer by radiologists. It is important to realize that mammographic image analysis is an
extremely challenging task for a number of reasons. As the efficacy of CAD systems can have
very serious implications, there is a need for perfection. Then, the large variability in the
appearance of abnormalities makes this a very difficult image analysis task. The abnormalities
are often occluded or hidden in dense breast tissue, which makes detection difficult.
1.3 Objective
Computer-aided methods are powerful tools that could assist medical staff in hospitals and lead
to better and more accurate diagnosis. Identifying representative, relevant and discriminant
image features for analysis and proper image classification. In the proposed system, Gabor
wavelets based features are extracted from medical mammogram images. On the extracted
features Principal Component Analysis (PCA) is further employed to reduce data dimensionality.
At the end, directional properties and frequency spectrum of those features are analyzed
with respect to the classification performance by employing support vector machines as
classifier. The results obtained indicate that Gabor wavelets provided by their orientation are
important issues to accurately discriminate mammogram tumor types. The proposed system
focuses on the solution of two problems. One is how to detect tumors as suspicious regions with
a very weak contrast to their background and another is how to extract features which categorize
tumors
The main objective of the research is to develop a CAD (Computer Aided Diagnosis)
system for finding the tumors in the mammographic images and classifies the tumors as Benign
or Malignant. There are five main phases involved in the proposed CAD system. They are image
pre-processing, extraction of features from mammographic images using Gabor Wavelet and
8
DWT (Discrete Wavelet Transform), dimensionality reduction using PCA and classification
using Support Vector Machine (SVM) classifier. Initially Image Preprocessing is done by
applying two dimensional median filter and then histogram equalization so as to get more
enhanced image. Then Gabor features and DWT features are extracted from the images which
are reduced by Principal Component Analysis. Further Support Vector Machine (SVM) classifier
is used to classify the tumor as Benign or Malignant or Normal.
9
CHAPTER 2 : LITERATURE SURVEY
There is an extensive literature on the development and evaluation of CAD systems in
mammography. Important visual clues of breast cancer include calcification clusters and
preliminary signs of masses. In the early stages of breast cancer, the signs are very subtle and
varied in appearance, making diagnosis difficult and challenging even for specialists. To decide
that suspicious area is malignant or benign, the tissue has to be removed for examination using
breast biopsy techniques. A false positive detection may result into unnecessary biopsy. In a false
negative detection, an actual tumor remains undetected that may lead to higher costs or even to
the cost of a human life. In addition, the tumors existing are of different types. Tumors are of
different shapes and some of them have the characteristics of the normal tissue. The American
Cancer Society estimates that 182,460 women in the United States will be found to have invasive
breast cancer in 2008. [1]. Pelin Gorgel, Ahmet Sertbas et al. [2] proposed an approach for
classification of mammographic masses as benign or malign. It uses Support Vector Machine
(SVM) and wavelet-based sub-band image decomposition. It uses two methods as feature
extraction by computing the wavelet coefficients and then classification using the classifier
trained on the extracted features. SVM was trained through supervised learning to classify
masses. The research involved 66 digitized mammographic images. The masses were segmented
manually by radiologists. The preliminary test on mammogram had shown over 84.8%
classification accuracy by using the SVM with Radial Basis Function (RBF) kernel. The Discrete
Wavelet Transform (DWT) is applied to each dimension separately [3].
N. Riyahi Alam, F. Younesi et al. [4] developed Novel hybrid segmentation method for
detection of masses in digitized mammograms using three approaches: Adaptive thresholding
method, Gabor filtering and fuzzy entropy feature as a computer-aided detection (CAD) scheme.
The proposed method was tested on 78 mammograms from the BIRADS databases. The detected
regions were confirmed by comparing them with the radiologist's hand-sketched boundaries of
real masses. This algorithm can achieve a sensitivity of 90.73% and specificity of 89.17%. M.
Vasantha, Dr. V. Subbiah et al. [5] have proposed a hybrid approach of feature selection is which
reduces 75% of the features. Decision tree algorithms are applied for classification of
mammography images by using these reduced features. This technique of classification was not
implemented before and it reveals the potential of Data mining in medical treatment. It uses
10
contrast limited adaptive histogram equalization (CLAHE) method is used for reducing the noise
produced in homogeneous areas and was originally developed for medical imaging [6]. This
method was used for enhancement to remove the noise of digital mammogram [7]. CLAHE
operates on tiny regions in the image called tiles rather than the entire image. Each tile's contrast
is enhanced, so that the histogram of the output region approximately matches the uniform
distribution or exponential distribution. The experimental results of enhancement on digital
mammogram using CLAHE have been reported [8]. In analyzing mammogram image [9], it is
important to distinguish the suspicious region from its surroundings. The methods used to
separate the region of interest from the image and dividing an image into distinct, meaningful
regions is called image segmentation. The benefit of this method is it requires no training data or
prior knowledge of the image contents.
The study in [10] by Nalini Singh et al. shows the outcome of applying image processing
threshold, edge based and watershed segmentation on mammogram breast cancer image and a
case study between them based on time consuming and simplicity. J. Subash Chandra Bose, K.
R. Shankar Kumar et. al [11] presents a new method for detection and classification of micro
calcifications. It uses four stages: first, preprocessing stage deals with noise removal, and
normalized the image. Second, Fuzzy c-Means clustering (FCM) is used for segmentation and
pectoral muscle extraction using area calculation and then micro calcifications detection. The
third stage two dimensional discrete wavelet transform is extracted from the detection of micro
calcifications. Finally, the extracted features are passed to the Artificial Neural Network that is
further classified into normal or abnormal images. Matsubara et al. [12] developed an adaptive
thresholding technique that uses histogram analysis to divide mammographic image into
different categories based on the density of the tissue from fatty to dense.
Nawazish Naveed et al. [13] proposed a technique to enhance the classification of
mammograms using multi-classification six abnormality classes as ill-defined masses, Well-
defined/circumscribed masses, Calcification, Speculated masses, Architectural distortion,
Asymmetry and Normal. The system is developed for diagnosing the breast cancer from
mammogram images. Preprocessing on mammogram image is performed to minimize the
computational cost and maximize the probability of accuracy. In second phase Discrete Wavelet
Transform (DWT) features are extracted for classification of mammogram into malignant and
benign. Later, the malignant images are again classified using one against all technique to find
11
abnormalities present in the mammograms. It has achieved average accuracy of classification
97.45% in detection of malignant and benign mammograms from MIAS dataset. Ioan Buciu, A.
Gacsadi [14] proposed Gabor Wavelet based feature extraction for Medical Image Analysis and
Classification. In this paper, Gabor wavelets based features are extracted from medical
mammogram images representing normal, or benign and malign tumors. Further, Principal
Component Analysis (PCA) is applied to reduce data dimensionality. At the end, directional
properties and frequency spectrum of the features are analyzed with respect to the classification
performance by applying multiclass support vector machines classifier.
2.1 What is Mammography?
Mammography is a special type of x-ray imaging that produces detailed images of the breast.
According to US Food and Drug Administration (FDA) report there are about 33.5 million
mammography performed per year in the United States. [15]. Mammography can show changes
in the breast before a woman can feel them. If a breast abnormality is found or confirmed with
mammography. There are two types of mammography: (1) Screening (2) Diagnostic. Screening
mammography is an x-ray examination of the breasts in a woman who has no symptoms of
breast cancer. The Screening mammography detects cancer at the earliest stage before its
symptoms noticed by physician. Early detection of small breast cancers by screening
mammography improves a woman's chances for successful treatment [16]. Screening
mammography is recommended every one to two years for women above 40 years of age. The
screening mammography is less expensive than diagnostic mammography. Women with
personal history of breast a cancer or breast implants may be needed to do diagnostic
mammography.
2.1.1 What is a Mammogram?
A mammogram is an x-ray exam of the breast used to evaluate breast changes. X-rays were first
used to examine breast tissue, by the German surgeon, Albert Salomon. Mammograms today
expose the breast to much less radiation compared with those in the past. A mammogram may
show something suspicious, but by itself it can't prove that an abnormal area is cancer [16]. If a
mammogram raises a symptom of cancer, a tissue sample from the suspicious area is removed
12
and examined under the microscope to find out if it's cancer. The doctor reading the
mammogram will look for different types of changes.
Calcifications
The tiny mineral deposits within the breast tissue that look like small white spots on a
mammogram are called calcifications. It may or may not be caused by cancer. There are 2 types
of calcifications.
Macro-calcification
The coarse (larger) calcium deposits that are most likely due to changes in the breasts caused by
aging of the breast arteries, inflammation, or old injuries are called Macro- calcification. These
deposits are non-cancerous conditions and do not require a biopsy.
Micro-calcifications
The tiny specks of calcium in the breast are called Micro-calcifications. The presence of micro-
calcifications does not mean that cancer is present. The layout and shape of micro-calcifications
help the radiologist to detect cancer. In most cases, the presence of micro-calcifications does not
mean a biopsy is needed. But if the micro-calcifications have a suspicious look and pattern, a
biopsy is recommended. During a biopsy, a small piece of the suspicious area is removed to be
looked at under a microscope.
A Mass or Cyst
A mass, with or without calcifications, is another important change seen on a mammogram.
Masses are areas that look abnormal and they can be cysts and non-cancerous solid tumors such
as fibroadenomas. The simple fluid-filled sacs are known as simple cysts and partially solid
known as complex cysts. Simple cysts are benign and don't need to be biopsied. Solid tumor or
complex cyst need to be biopsied to be sure it isn't cancer. A tumor and a cyst can feel same on a
physical examination and on a mammogram. To confirm that a mass is really a cyst, a breast
ultrasound is often done.
13
2.1.2 Limitations of Mammograms
Although breast cancer screening is the best way to find cancer early does not always reduce a
woman's mortality rate due to breast cancer. Even though mammograms can detect breast
cancers too small to be felt, treating a small tumor does not always mean it can be cured. ACS
[16] screening guidelines suggest that women with serious health problems should ask their
doctors whether to continue having mammograms.
False-negative results
A false-negative mammogram seems normal even though breast cancer is present. The screening
mammograms may miss about 1 in 5 breast cancers. The main reason of false-negative results is
high breast density. It occurs more often among younger women than older women because
younger women may have more dense breasts. Breasts usually become less dense as women age.
False-positive results
A false-positive mammogram looks abnormal but cancer is not present. Abnormal mammograms
require extra testing that includes ultrasound, diagnostic mammograms, and sometimes biopsy.
False-positive results are more common in women who are younger, have dense breasts, have
breast cancer in the family, have had breast biopsies, or are taking estrogen.
2.1.3 How is Mammography Performed?
For mammography, the technologist may position the patient and take image of each breast
separately. One at a time, the breasts are carefully positioned on a special film cassette and then
gently compressed between 2 paddles attached to the mammogram machine unit to spread the
tissue distant. This squeezing or compression ensures that there will be very little movement to
get the sharper image. The examination can be done with a lower x-ray dose by flattening the
breast in order to get the maximum amount of tissue to be imaged and examined. Although this
compression may be uncomfortable, it lasts for a few seconds and is necessary to produce a good
mammogram. The entire procedure for a mammogram takes about 20 minutes.
Some of the mammography technologists may place adhesive markers to the breast skin
prior to taking images of the breast. The purpose of the adhesive markers is: first, to identify
areas with moles, blemishes or scars so that they are not falsely identified as an abnormalities,
and secondly, to identify areas that may be of concern e.g. a lump was identified during physical
14
examination. To provide a clear landmark for the radiologist on the mammogram images, some
centers mark the nipple with a small dot. To get a mammogram, the x-ray source is turned on and
x-rays are passed through the compressed breast and cassette. The x-rays strike on a special
phosphor coating inside the cassette. Depending upon the hitting intensity of the x-ray beams, the
phosphor glows proportionally. To create the high quality images at the lowest exposure highly
sensitive film and special x-rays are used for mammography. The exposed film inside the
cassette is then developed in a dark room like a photograph is developed. The special energy and
wavelength of the x-rays allow them to pass through the breast and create the image of the
internal structures of the breast. When the x-rays pass through the breast, they are attenuated by
the different tissue densities they come across. The connective tissue over the breast ducts and fat
is denser and absorbs less x-ray energy. The differences in absorption and the varying exposure
level of the film create the images which can clearly show normal structures such as fat, breast
ducts, fibro-glandular tissue, and nipples. Further, abnormalities such as micro-calcifications,
cysts and masses are also visible. The developed mammography films are interpreted by a
radiologist, who compares the new images of a woman's breast to older mammograms a woman
has had. The radiologist will look for shadows and patterns of tissue density to detect any
abnormalities.
A mammogram is similar to a fingerprint. The mammogram varies extremely from
woman to woman, and no two mammograms are similar. It is very helpful for the radiologist to
have films available from previous examinations for comparison to identify small changes that
occur gradually over time and detect a cancer as early as possible. The breasts of an adult woman
are tear-shaped and milk producing glands. They are attached to the front of the chest wall on
both side of the breast bone or sternum by ligaments. The muscle tissue is not present in breast.
The glands are surrounded by a layer of fat that are further extended throughout the breast.
Details
- Pages
- Type of Edition
- Erstausgabe
- Year
- 2017
- ISBN (PDF)
- 9783960676416
- ISBN (Softcover)
- 9783960671411
- File size
- 3.2 MB
- Language
- English
- Institution / College
- Dr. Babasaheb Ambedkar Marathwada University
- Publication date
- 2017 (March)
- Grade
- 8.00
- Keywords
- Mamogram Gabor wavelet Mammography Mastography X-rays Diagnostics Diagnosis Breast cancer Precautions against cancer Precautions against breast cancer Wavelet analysis Support Vector Machine Matlab Discrete Wavelet Transform Computer Aided Diagnosis CAD Principal Component Analysis
