A Minor Project Report

On

IDENTIFICATION OF BREAST CANCER BY USING

MATLAB

Submitted in partial fulfilment of requirements for the award of the

Degree of

BACHELOR OF ENGINEERING

in
ELECTRONIC AND COMMUNICATION ENINEERING

Under the guidance of

Supervisor Name

Ms. JANANI M

Submitted By

NAVEEN V(18BEC4107)

SATHISH KUMAR S(18BEC4149)

SURIYA M(18BEC4172)

YAATHASH B(18BEC4193)

DEPARTMENTOF ELECTRONIC AND COMMUNICATION ENGINEERING

M.KUMARASAMY COLLEGE OF ENGINEERING
(Autonomous)
KARUR – 639 113
SEPTEMBER,2020
 BONAFIDE CERTIFICATE

Certified that this Minor project – II report Identification Of Breast Cancer By

Using MATLAB is the bonafide work of Naveen V (18BEC4107), Sathish Kumar S
(18BEC4149), Suriya M (18BEC4172), Yaathash B (18BEC4193) who carried out the
project work under my supervision in the academic year 2019-2020.

Signature of the Students:

1.Naveen V

2.Sathish Kumar S

3.Suriya M

4.Yaathash B

Signature of the Guide Signature HoD

 Vision and Mission of the Institute

Vision

To emerge as a leader among the top institutions in the field of technical education

Mission

M1: Produce smart technocrats with empirical knowledge who can surmount the global
challenges.
M2: Create a diverse, fully-engaged, learner-centric campus environment to provide quality
education to the students.
M3: Maintain mutually beneficial partnerships with our alumni, industry and professional
associations
Vision and Mission of the Department
Vision
To empower the Electronics and Communication Engineering students with emerging
technologies, professionalism, innovative research and social responsibility.
Mission
M1: Attain the academic excellence through innovative teaching learning process, research
areas & laboratories and Consultancy projects.
M2: Inculcate the students in problem solving and lifelong learning ability.
M3: Provide entrepreneurial skills and leadership qualities.
M4: Render the technical knowledge and industrial skills of faculties.
 PROGRAM EDUCATIONAL OBJECTIVES (PEOs)
PEO 1: Graduates will have successful career in software industries and R&D divisions through
continuous learning.
PEO 2: Graduates will provide effective solutions for real world problems in the key domain of
computer science and engineering and engage in lifelong learning.
PEO 3: Graduates will excel in their profession by being ethically and socially responsible.

PROGRAM OUTCOMES (Pos)

Engineering students will be able to:

1. Engineering knowledge: Apply the knowledge of mathematics, science,

engineering fundamentals, and an engineering specialization to the solution of
complex engineering problems.
2. Problem analysis: Identify, formulate, review research literature, and analyze
complex engineering problems reaching substantiated conclusions using first
principles of mathematics, natural sciences, and engineering sciences.
3. Design/development of solutions: Design solutions for complex engineering
problems and design system components or processes that meet the specified needs
with appropriate consideration for the public health and safety, and the cultural,
societal, and environmental considerations.
4. Conduct investigations of complex problems: Use research-based knowledge and
research methods including design of experiments, analysis and interpretation of
data, and synthesis of the information to provide valid conclusions.
5. Modern tool usage: Create, select, and apply appropriate techniques, resources, and
modern engineering and IT tools including prediction and modeling to complex
engineering activities with an understanding of the limitations.
6. The engineer and society: Apply reasoning informed by the contextual knowledge to assess
societal, health, safety, legal and cultural issues and the consequent responsibilities relevant
to the professional engineering practice.
7. Environment and sustainability: Understand the impact of the professional
engineering solutions in societal and environmental contexts, and demonstrate the
knowledge of, and need for sustainable development.
8. Ethics: Apply ethical principles and commit to professional ethics and
responsibilities and norms of the engineering practice.
9. Individual and team work: Function effectively as an individual, and as a member
or leader in diverse teams, and in multidisciplinary settings.
 10. Communication: Communicate effectively on complex engineering activities with
the engineering community and with society at large, such as, being able to
comprehend and write effective reports and design documentation, make effective
presentations, and give and receive clear instructions.
11. Project management and finance: Demonstrate knowledge and understanding of
the engineering and management principles and apply these to one’s own work, as a
member and leader in a team, to manage projects and in multidisciplinary
environments.
12. Life-long learning: Recognize the need for, and have the preparation and ability to
engage in independent and life-long learning in the broadest context of technological
change.
PROGRAM SPECIFIC OUTCOMES (PSOs)

PSO1: Professional Skills: Ability to apply the knowledge of computing

techniques to design and develop computerized solutions for the problems.
PSO2: Successful career: Ability to utilize the computing skills and ethical values
in creating a successful career.

MAPPING OF PROJECT WITH POs and PSOs

Title COs
CO1 CO2 CO3 CO4 CO5

POs
Minor PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10
Project – 1I
PSOs
PSO1 PSO2
 ABSTRACT

Breast cancer is one of the most common cancer affecting women around the
world. Mammography is the most known and effective method to detect early signs
of breast cancer. However, due to some weaknesses in mammography such as
painful procedures and radiation, researches introduce another diagnosis method
which is by analysing thermal image. The purpose of this project is to design a
system to detect the signs shown in mammogram and thermal image using Image
Processing Technique applied to MATLAB. Image processing techniques can be
divided into several elements. The elements are image acquisition, image pre-
processing, image processing, feature extraction, object classification and
classification decision. Both type of images then analyse step by step according to
the elements. Mammogram images are analyse using morphology technique before
features extraction which then lead to the classification of the image into 3 classes
(„Normal Fatty breast‟, „Abnormal Fatty breast‟ and „Glandular breast). For thermal
image, the distribution of heat around the breast will be the features extracted and
analysed. The different range of heat in the image will be used to specify the possible
area of cancer. This project also includes the construction of Graphical User Interface
(GUI) so that the system is more users friendly.
 TABLE OF CONTENT

CHAPTER TITLE PAGE

1 INTRODUCTION 1

1.1 Problem Statement 2

1.2 Objectives 2
1.2.1 Sub-Objective 3

1.3 Project Scope 3

2 LITERATURE REVIEW 4

2.1 Image Processing Technique 4

2.2 Mammogram Image Analysis 6
2.3 CAD system using Thermal Image 8

3 RESEARCH METHODOLOGY 10

3.1 Image Acquisition 11

3.1.1 System 1 11
3.1.2 System 2 12
3.1.3 System 3 12

3.2 Image Pre-processing and Image Processing 14

3.2.1 System 1 14
3.2.2 System 2 15

4 CONCLUSION 16

4.1 Result 16

4.2 Future Recommendation 16

REFERENCE 17
 LIST OF TABLES

TABLE NO. TITLE PAGE

3.1 Sobel approximation to the derivatives. 15

LIST OF FIGURES

FIGURE NO. TITLE PAGE

1.1 Breast cancer. 1

3.1 Image processing main elements. 10

3.2 YSX0906 Digital Mammography X-ray Machine. 11

3.3 Mammogram image. 12

3.4 FLIR A615 infrared camera. 13

3.5 Thermal image of breast. 13

3.6 Image Pre-processing and Image Processing. 14

3.7 Image Pre-processing and Image Processing. 15

 CHAPTER 1

INTRODUCTION

Breast can be classified into 2 types due to its density which are „Fatty breast‟
and „Glandular breast‟. When the amount of fat tissues exceed the amount of fibro-
glandular tissues, the breast can be classified as „Fatty breast‟ and when the amount
of fibro-glandular tissues exceed the amount of fatty tissues, the breast can be
classified as „Glandular breast‟. Breast cancer occurs when breast tissues grow,
change and multiply rapidly without control which may form lump or mass of extra
tissues as shown in Figure 1.1. These masses are called tumor and can be either
cancerous (malignant) or non-cancerous (benign) [1].

Figure 1.1 Breast cancer

Breast cancer is one of the most common cancers affecting woman and the
most common source of death among middle aged women. Based on the World
Health Statistics 2011 by Global Health Observatory (GHO), the mortality among
female population all over the world cause by malignant neoplasma is about 11.81%
and breast cancer are the highest with 15.80% compare to other types of cancer [2].
Successful treatment of breast cancer depends on early detection. Currently, two
imaging method uses to detect masses are mammography and thermography.
 2

1.1 Problem Statement

Mammography is a specific type of x-ray imaging that focusing on breast

imaging. This process uses system with low dose x-ray, high contrast and high-
resolution film [3]. However the accuracy in detecting the cancer based on
mammogram image with bare eye by qualified personnel (radiologist) will be
affected by poor mammographic image quality and fatigue radiologist. Radiologist
misdiagnose 10-30% of the malignant cases due to the difficulty to maintain required
attention level when reading large number of screening mammograms since most are
free of cancerous features [1]. However, due to the invasive process of
mammography, another method emerged as potential method to detect the early sign
of breast cancer. The method is thermography.

Thermography is an imaging technique which shows temperature patterns at

the surface of the breast. The heat patterns indicate the metabolic activity and
vascular circulation in breast tissue which will be high at the surrounding area of
cancerous tissue [4].

Due to the need of overcoming the problem that cause high rate of false
positive and false negative detection, a Computer Assisted Detection (CAD) system
is develop to provide assistant for clinician to identify cancerous tissue in
mammogram and thermal image. The system will be design based on image
processing technique on MATLAB platform.

1.2 Objectives

This project proposes to develop a system for breast cancer using image
processing technique.
 3

1.2.1 Sub-objective

The sub-objective for this project is as follow.

1. To classify type of breast and detect abnormalities of breast using

mammogram image.

2. To study thermal image processing feasible to detect breast cancer.

1.3 Project Scope

The 150 breast mammogram images for this study are obtained from trusted
online database (MIAS database) [11]. Breast thermal image with abnormalities
obtain from 6 case studies by Pacific Chiropractic and Research Centre Infrared
Imaging in California [4]. Both type of image analyse using Matlab software.
 4

CHAPTER 2

LITERATURE REVIEW

Several researches have been done to develop CAD system to detect breast
cancer. The references for this paper are taken from journal, books and conferences
regarding the mammogram image and thermal image.

2.1 Image Processing Technique

Alasdair McAndrew (2004) entitlement, image processing is used to change

the nature of an image to improve and enhance the image for human interpretation.
Image processing also used to render image for machine perception. In his module,
he explains on how to use matrix capabilities of MATLAB to investigate images and
its properties. The image processing operations are explained in term of chapters.
Image display chapter explain on how to use imshow function to display image and
how spatial resolution and quantization affect the display and appearance of the
image. Another chapter explain on point processing and the sub-chapter are
arithmetic operation, histogram and thresholding. This sub-chapter discuss on how to
modify image (enhance and blurring image) using MATLAB function that show how
each operation work. For example, imadjust function which indicate histogram
stretching is use to enhance image. The next chapter teach about spatial filtering.
This chapter is also explaining on how to enhance and blurring image but using
different operation. The operation discuss in this chapter is by filtering image using
frequencies (low and high pass filter), Gaussian filter and non-linear filter. Types of
noise, cleaning salt and pepper noise and cleaning Gaussian noise are explain in the
 5

next chapter which is noise chapter. Noise is degradation in the image cause by
image disturbance during transferring and during image acquisition. Cleaning noise
is important to restore image to its original state and to analyse the image. The type
of noise discuss in this module are salt and pepper noise and Gaussian noise. The
image is filtered using fspecial function to clean up the noise. The next chapter
explain one of the most useful information in an image which is edge. The uses of
finding the edge are to measure size of the object, to isolate object from background,
to recognise and classify object in the image. There are numbers of edge detection
method discuss in this module such as Robert edge detection, Sobel edge detection
and Prewitt edge detection. The module also discuss on topic morphology.
Morphology is an operation in image processing to analyse shape in image.
Morphology consist of many types of operations and some of it such as dilation,
erosion, opening, closing, hit or miss transform, region filing and connected
components are discuss in this module. Topic colour processing is discussed in the
next chapter. In this chapter, the main topic discuss are what is colour mean in image
processing, colour models, colour image in MATLAB, pseudo colouring and colour
images processing . For example, to extract RGB component in RGB image, imshow
function can be used. [17]

Marius Leordeanu et. al. (2011) entitlement, boundary detection is an

important task when doing segmentation and recognition using vision system.
Despite being one of the most important tasks to be done in image processing, there
is no general formulation for boundary detection. This paper discusses on
formulation and algorithm designed to detect different types of boundaries such as
boundaries intensity, occlusion boundaries and specific boundaries for object. Based
on their observation, boundary can be summarizing as a region that separates
different image regions or a layer that coincide with boundaries in other layer. In this
paper, author has designed two algorithms (Gb1 and GB2) and tested the algorithm
to detect boundaries in static colour images, occlusion boundaries in video, occlusion
boundaries in RGB-D video and boundaries from soft-segmentation. The result is the
algorithm effectively and accurately detects boundaries of images use in the
experiment. [18]

Hao Yuen Kueh et. al. entitlement, biological image contain a lot of patterns
and objects which may convey information about biology mechanism. This tutorial
 6

discusses the process to extract data from raw microscopic image using MATLAB.
The advantages to extract and quantify objects and patterns using automated image
analysis compare to manual methods of analysis is automated image analysis will
provide unbiased approach to extract information from images and testing
hypotheses. Automated system analysis also has advantages to facilitate the
collections of large data collections for statistical analysis. The topics discuss in the
first section of this tutorial are how to read, display, write and convert images.
Besides that, the author also discuss on how MATLAB represent image and how to
convert between different types of image. The second section discuss on contrast
adjustment. As majority of biological image have low dynamic range and the
features are difficult to be analysis, there is a need to enhance the appearance of the
image by using different intensity transformation. This step may improve the
performance of image segmentation algorithm and feature recognition. Next section
discuss on spatial filtering technique. The filters explained are smoothing filters
(average filter and Gaussian filter), edge detection filter (Prewitt filter and Sobel
filter), Laplacian filter and median filter. Mathematical morphology which uses to
extract features and components in images discussed on the next chapter. The
operations are dilation, erosion, opening (erosion followed by dilation), closing
(dilation followed by erosion), filling holes and clearing border objects. Image
segmentation process to subdivide image into regions and images discuss in the next
section. The quantitative information is processed and analysed using segmentation
technique for extraction. The techniques of segmentation are edge detection and
morphological watershed. This tutorial also discuss on analysis of dynamic and
motion in biological images. The techniques use to visualize dynamical behaviour
are kymographs (two-dimensional analog of times traces), difference images,
maximum intensity projections, image cross-correlation and particle tracking. [19]

2.2 Mammogram Image Analysis

Jelene Bozek, Kresimir Delac and Mislav Grgic (2008) entitlement,

mammography is the best method to detect early signs of breast cancer such as
masses, calcifications, bilateral asymmetry and architectural distortion. However,
 7

due to human limitation computer system have to take the major role in detecting
abnormal tissue. The challenges that have been faced by the system are the wide
range of abnormalities features and the indistinguishable from surrounding cell. Most
of system developed involves algorithms which consist of two stages. The first stage
is to detect suspicious lesion and second stage is to reduce the number of false
positives. In BI-RADS system which discussed in this paper, the detected lesions are
classified as masses, calcifications, architectural distortion and bilateral asymmetry.
Masses are classified as benign or malignant based on density (fat containing, low
density, isodense and high density), margins (circumscribed, microlobular, obscured,
indistinct and spiculated) and shape (round, oval, lobular and irregular).
Calcifications classified as benign, malignancy suspicious and malignancy highly
suspicious based on the distribution of cluster, size, shape, and variability. For
architectural distortion, the lesion classified as malignant when integrated with other
lesion such as masses and classified as benign when scar and self-tissue damage due
to trauma detected. Bilateral asymmetry analyse based on its texture, shape
measurement, topology, brightness distribution, roughness, pattern assymetry and
directionality. [3]

Ranjeet Singh Tomar et. al. (2009) entitlement that image processing
techniques that been mention in their journal are more radiologist friendly. The
system is designed using image processing technique on MATLAB. The techniques
uses are edge detection and morphology. The detection process designed will start
with detecting entire cell in the image, followed by filling gaps, dilating gaps,
removing border, smoothing the objects, finding structures and lastly extracting large
objects. For feature extraction to find the wanted area, 3 steps were suggested which
were reduce uneven illumination, determine size distribution in Top-hat Image and
calculate first derivatives. The result for feature extraction then plotted into graph to
be analysing for classification. [6]

Hala Al-Shamlan and Ali El-Zaart (2010) entitlement, the features extraction
in mammogram is an important key for early detection of breast cancer. In this study,
they aim to determine the features extraction range. Before the range was
determined, image pre-processing and image segmentation process applied to the
images. Image pre-processing done to the image to suppress noise and improve the
contrast of the image. Image segmentation is for detect the suspicious lesion. The
 8

features used are based on three main categories; Geometric, Texture and Gradient
features. For Geometric category, the features measured are area, perimeter and
compactness [20]. Features in Texture category mostly obtain from image histogram.
The features are mean, mean global area, mean local area, uniformity, standard
deviations, smoothness, skewness, entropy, correlation and inverse. The last category
is Gradient category. Features that classified under this category are Sobel-mean,
Sobel-mean global area, Sobel-mean local area, Sobel-uniformity, Sobel-standard
deviations, Sobel-smoothness, Sobel-skewness, Sobel-entropy, Sobel-correlation and
Sobel-inverse. After applying up to 23 types of features extraction to 80
mammograms, they manage to obtain the range value for each feature extraction
which may be used for further process in their breast cancer CAD system. [7]

2.3 CAD System Using Thermal Image

Monique Frize, Christophe Herry and Rober Roberge (2002) entitlement, the
3 technique in Head et al‟s methods [21]. The study shows the third method provided
reliable result compare to the first and second method when applied to 9 patient‟s
sample (6 with a diagnosis of normal and 3 with cancer). One of the analysis done is
by increasing the threshold value in the methods and the result obtain are no false
negatives or false positives on the sample. Therefore by looking at this preliminary
result, they concluded future work should focus on improving third method to
enhance thermogram diagnosis and decrease false negatives or false positives. [8]

V. Umadevi, S. V. Raghavan and Dr. Sandeep Jaipurkar (2010) entitlement,

an interpretation system able to characterize thermography image as normal or
required follow-up with clinician. The infrared cameras uses in this paper are Ti40FT
from M/s Fluke Corp. and Varioscan-3021 ST from Jenoptik Laser. Software that
integrated with the cameras to view images captured are SmartView for Ti40FT and
IRBIS for Varioscan 3021-ST. The system discussed in this paper is Infrared
Thermography Based Image Construction (ITBIC) system. This system consists of
two main process, body boundary identification and highest temperature area
extraction. This system will classify the image into normal case or follow-up case.
When the system tested to 50 female volunteers, the system manages to characterize
 9

and match the result to clinical finding. The system then interface with developed
graphical user interface (GUI) to allow easier thermal image analysis by the
radiologist or clinician. [9]

Pragati Kapoor and Dr. S. V. A. V. Prasad (2010) entitlement, image

segmentation with automatic approach may improve the accuracy in earlier detection
of breast cancer for thermography image. The methods outlined in this study are
image segmentation and asymmetry analysis. One of the image segmentation process
that use in this paper is edge detection which extract the boundaries of the breasts.
The process also involving Hough transforms to extract the lower breast boundaries.
Segmented classification done to classify each segmented pixel into a certain number
of clusters. Lastly, diagnostic based on asymmetric analysing of the pixels in every
cluster. The features uses in this paper for the diagnosis of image are skewness and
kurtosis. [10]
 10

CHAPTER 3

RESEARCH METHODOLOGY

This study will be using image processing main elements which are image
acquisition, image pre-processing, image processing, feature extraction, object
classification and classification decision as shown in Figure 3.1 for developing CAD
system for mammograms and thermal images.

Image Image Pre- Image Feature Object Classification

Acquisition processing Processing Extraction Classification Decision

Figure 3.1 Image processing main elements

Image acquisition step involves the camera and its connection to the
computer or processors. Computer or processors will receive the image in digital
format. Image pre-processing step is a step to improve and enhance the image for
processing step. Image processing step is a further step in analysing the image to
obtain desired object. A lot of image processing techniques can be used in this step
such as morphological processing, edge detection and compression. Feature
extraction is where a set of desired features extracted from data pixels of the image
which are good for classification. Object classification and classification decision are
steps to make decision based on test and analysis done on the image [12]. The above
methods is use to develop the following system;

1. System 1: Classification between „Fatty breast‟ and „Glandular breast‟.

2. System 2: Detection for abnormalities in type „Fatty breast‟ images.
3. System 3: Detection of high temperature in thermal images.
 11

3.1 Image Acquisition

3.1.1 System 1.

Mammogram image is obtained using mammogram machine. There are many

types of mammogram machine and one of it is „YSX0906 Digital Mammography X-
ray System‟ manufactured by YSENMED from China as shown in Figure 3.2. The
sizes of mammogram image that can be obtained using this machine are 18 x 24 cm
or 24 x 32 cm. The C-arm of the machine can rotate between range +180° and -135°.
The clinician is allowed to magnify the image with ratio from (1.4)/1 until (1.6)/1.
The thickness of breast that can be compress from 0 mm until 268 mm with pressure
range between 0 kg until 20 kg [13].

Figure 3.2 YSX0906 Digital Mammography X-ray Machine

In this study, The CAD system will be tested using 150 mammography
images (65 „Fatty breast‟ image and 85 „Glandular breast‟ image). The digital
mammography images acquired from online mammogram database (MAIS
database). The image resolution of the image is 1024 x 1024 and in PGM (Portable
Graymap) format. A sample of the database is shown in Figure 3.3.
 12

Figure 3.3 Mammogram image

3.1.2 System 2.

For detection for abnormalities, the system will test the 65 „Fatty breast‟
mammography image that had been classified among the 150 mammogram images
tested before. The digital mammography images acquired from online mammogram
database (MAIS database). The image resolution of the image is 1024 x 1024 and in
PGM (Portable Graymap) format.

3.1.3 System 3.

Thermal images of breast can be obtained using infrared camera. One type of
camera that can be used is FLIR A615 manufactured by FLIR Systems, Inc. as
shown in Figure 3.4. FLIR A615 is a perfect instrument for industries when the
temperature changes over time is quiet fast. FLIR A615 also complies with standards
like GigE Vision that allow this camera to interface using the Gigabit Ethernet
communication protocol and fast image transfer using low cost standard cables even
over long distances. This camera also complies with GenICam protocol which allows
the camera to be use with third party software. Due to its compliance to standards,
FLIR A615 is a Plug&Play device within 3rd parties Machine Vision softwares like
NIs IMAQ Vision™ and the MVTecs Halcon™ software. By using this camera,
image with resolution 640 x 480 pixels can be obtained [14].
 13

Figure 3.4 FLIR A615 infrared camera.

For thermal image analysis, the system will test the 6 image at front position
that taken from 6 case studies by Pacific Chiropractic and Research Centre Infrared
Imaging website. The image resolution of the image is around 300 x 200 and in JPG
(Joint Photographic Expert Group) format. A sample of thermal image is shown in
Figure 3.5.1.

Figure 3.5 Thermal image of breast

 14

3.2 Image Pre-processing and Image Processing

3.2.1 System 1.

(a) (b)

(c) (d)

Figure 3.6 Image Pre-processing and Image Processing. (a) original image, (b)
BW image, (c) Sobel gradient of (b), (d) Cropped image based on Hough parameter.

The original image (Figure 3.5.2(a)) is a grayscale image. The image then
convert to BW (black and white) image (Figure 3.5.2(b)) using im2bw function.
im2bw (I, level) function converts grayscale image to a binary image. The output
image occur after replacing all pixels in the input image with luminance greater than
level with the value 1 (white) and replaces all other pixels with the value 0 (black).
The level for this function is between 0 and 1 which is relative to the signal levels
possible for the image's class.

Then, the edge of the image is finding by using Sobel gradient as shown in
Table 3.1. After Sobel gradient apply to image in Figure 3.5.2(b), image in Figure
3.5.2(c) obtained. By using this image, the parameters of Hough function and
Houghpeaks function obtain in order to crop the image automatically and produce
output image (Figure 3.5.2(d)). Hough (BW) computes the Standard Hough
Transform (SHT) of the binary image. Hough function is used to detect lines in the
image and generate Hough transform matrix. Houghpeaks function will locate peaks
in the Hough transform matrix and this value will be used to crop the ROI (Region of
Interest) of the image.
 15

-1 -2 -1 -1 0 1
0 0 0 -2 0 2
1 2 1 -1 0 1
Table 3.1: Sobel approximation to the derivatives

3.2.2 System 2.

(a) (b) (c)

(d) (e) (f)

Figure 3.7 Image Pre-processing and Image Processing. (a) original image, (b)
BW image, (c) imclearborder of image (b), (d) result after Morphological opening
apply on image (c), (e) output image after multiplication between image (a) and (d),
(f) sharpened image obtained by top-hat filtering and adjusting.

The image is crop to obtain new image with resolution 1001 x 1001 pixels.
The images then change to BW using im2bw function (Figure 3.6(b)) before the
noise removal function (imclearborder) use on the image (Figure 3.6(c)).
Imclearborder function use to suppress structures that are lighter than their
surroundings and that are connected to the image border. This step is to allow more
accurate further analysis.
 16
CHAPTER 4

CONCLUSION

4.1 Result

In pre-processing stage, the images read and saved as gray scale. Many filters were
applied such as white and black filter, sharp image and log transformation. In this
stage, the noise was removed followed by sharpening of the images and median filter
application. The median and log transformation give best results comparing with the
others. They work through mathematic equations to get rid the unwanted values of
pixels and substituting them with wanted ones. They work in extemporized manner
(non-linear) that increases the diagnostic and quality values of the images.

4.2 Future Recommendation

With additional features such as window means and standard deviations,

methods such as the k-nn algorithm are able to significantly reduce the
number of mislabeled pixels with respect to certain regions within the image.
In the future we plan to study the incorporation of additional region-based
features and speed-up techniques to improve the final segmentation. In
addition, we plan to include performance ratings from practicing clinicians
and comparison to DDSM ground truth to evaluate the efficacy of our final
system outputs.
 17
REFERENCES

 American Cancer Society (1999). Breast Cancer Facts & Figures 1997-
1998.

 Math Lab- mathworks.com

 Kopans, D. B. (1996). The potential impact of computer-aided diagnosis

on clinical mammography. Proceedings o] the Third International
Workshop on Digital Mammography, 35.